Read Thesis-Based Selling online: Winning Enterprise AI Deals

Page 1 of 214. People buy the thesis before they buy the product. THESISBASED SELLING. Winning enterprise AI deals before there are reference customers. FRU NDE

Page 2 of 214.

Page 3 of 214. Thesis-Based Selling. ™ Winning enterprise AI deals before there are reference customers. FRU NDE

Page 4 of 214. Thesis-Based Selling™: Winning Enterprise AI Deals Before There Are Reference Customers Copyright © 2026 Fru Nde. All rights reserved. You can read this book online at tbs.fru.dev. The workbook pages may be copied for use by your own team. Cite as: Fru Nde, Thesis-Based Selling, first edition, 2026, tbs.fru.dev. SPIN Selling, Strategic Selling, Solution Selling, MEDDIC, The Challenger Sale, Gap Selling, The JOLT Effect, Magic Quadrant, Forrester Wave and the other product and company names in this book are the property of their owners. This book is not affiliated with or endorsed by any of them. The stories in this book are illustrations. The companies and people in them are invented, built from the kinds of deals the author has worked on so that no real customer can be recognized. No story describes a real customer. The views are the author’s own and do not represent any employer. This book describes a way of working. It is not professional advice and promises no result. On the cover: a murmuration of starlings. Thousands of birds, a few simple rules, and a shape that nobody directs. First edition, 2026. Fru Nde

Page 5 of 214. C O N T E N TS F R O N T M AT T E R A Note Before You Start Introduction: The Turning Point 5 9 PA R T I · W H Y S E L L I N G H A S TO C H A N G E ONE TWO THREE FOUR The Proven Path, and Why It Is Breaking People Buy the Thesis Before the Product Selling When There Is No Reference Customer What Every Great Sales Method Quietly Assumes 25 37 51 65 PA R T I I · E S TA B L I S H T H E T H E S I S FIVE SIX SEVEN What a Thesis Is, and How to Test One Find the Beliefs Your Deal Rests On Five Theses About AI Inside Companies 85 105 115 PA R T I I I · S E L L O N T H E T H E S I S EIGHT NINE TEN ELEVEN Agree with Your Own Team Before the Customer Build the Sale Up from What Is True Defend the Thesis, and Let the Buyer Attack It Use AI to Attack Your Thesis Before the Buyer Does 133 143 157 165 B AC K M AT T E R Post Your Thesis Where Everyone Can Argue with It Field Guide: The Framework at Work Workbook: Three Steps, Twice Appendix: The Field Notes Appendix: List of Figures Resources: Further Reading and Exercises Notes: Sources by Chapter About the Author 173 177 189 199 201 203 207 211

Page 6 of 214.

Page 7 of 214. A Note Before You Start This book is built to be used as well as read. Read the first two parts straight through. Take the third into your next deal, a chapter at a time, with the Field Guide and the workbook at the back. It makes one argument. On a large decision, people buy the thesis before they buy any product. A thesis is a belief about how the world works, written down, agreed on and defended. Sellers mostly ignore it and sell features, and for a long time they could get away with it, because they had a second tool: somebody else’s success. The reference customer, the case study, the logo on the slide. In AI that tool is mostly missing, because almost nobody has run this technology long enough to be a reference. What is left is the thesis. You establish yours, get the buyer to agree on common ground, and build the sale on what follows. I call it the ThesisBased Selling framework. Why the name. Thesis-Based Selling, because people buy the thesis before they buy the product. Winning Enterprise AI Deals Before There Are Reference Customers, because that is the job now. At the pace this moves, 5

Page 8 of 214. Thesis-Based Selling most enterprise AI deals are won before a reference exists that matches the buyer’s case. So you win them on a thesis the buyer agrees with, not on proof that arrives too late. Who this is for. People who sell complex technology to large enter‐ prises. The sales engineer who has to stand in front of a customer and say what the product can do. The account executive forecasting a deal in a market that changes every quarter. The technical consultant, the solution architect and the founder who have to drive a customer to a decision nobody has made before. I write as a sales engineer, and every example is told from the seller’s side of the table. What this book is. Mostly synthesis, and I want to be plain about where it comes from. The idea that people buy a belief about the future before they buy a product is old. So are insight-led selling, the habit of challenging a buyer’s assumptions, and the study of how people decide under uncertainty. Selling has seventy years of good frameworks, and Chapter 4 names them and their authors. This book does not replace MEDDIC, SPIN or Chal‐ lenger. It gives them a foundation for making decisions when historical proof is not available. The problem it is for. For decades, sellers walked into the room knowing things buyers did not: how the product really behaved, which customers had made it work, what the market was doing, the lessons of hundreds of deals. That advantage is eroding, because of three shifts that arrived together. The first is the consumerized C-suite: the executives who sign now use the technology themselves, at home, and test your claims on their phones before the meeting (the Introduction). The second is the procurement time trap: a buying process that takes the best part of a year, for a technology whose AI models leapfrog each other in months (Chapters 1, 2 and 9). The third is the missing reference: the buyer asks who else is doing this, and the honest answer is that almost nobody has done it long enough to know (Chapter 3). Together they create what I call symmetric uncertainty: the buyer does not know where this market is going, and neither does the seller. I have not found a sales book written for that combination, so this one is. The seller’s job in it is not to pretend to have certainty. It is to bring a point of view you can defend, test it with the buyer, and build conviction together. How it is built. Every chapter opens with a card that has two lines on it: the chapter’s thesis, and what follows if the thesis holds. The book uses its 6

Page 9 of 214. Before You Start own method. If you disagree with a card, write down your reason. That is the exercise. The last section is a workbook: two rounds of three steps, each step with a page to fill in. Field notes. As you read you will come to numbered field notes, short lines set apart on the page. A field note is a lesson from the field: something practical I learned in front of a customer that you can take away and use. All of them are gathered in an appendix at the back. The stories in it. The stories are illustrations. Maya, Daniel, Priya and the companies in the examples are invented, built from the kinds of deals I have worked on, so that no real customer can be recognized. The research is real, and the notes at the back say where each number comes from. Before you read, write down your bets. Take a sheet of paper and put today’s date at the top. Then write five bets on where your industry and AI will be in two years, in five and in ten. Do not wait until you have finished the book, and do not wait until you are sure. A few questions to start you: What does a typical employee at your customer do by hand today that they will hand to AI agents? Which software category you sell into will not exist in its current form? What will your buyer believe in five years that they would reject today? What will your own job look like when the product can explain itself? Which of your customers will be the first to stop buying something they buy today? What will still be true no matter how good the AI models get? Keep the page. The workbook at the back has one for it. I do this myself, on quantum computing, on 6G and on much else. Most of the time I have been wrong. Rereading the pages is still one of the most useful things I do, because a dated bet shows me how I was thinking and sharpens the next one. No vendor is the hero. I name no product as the answer in this book. A thesis has to be true no matter whose logo is on the contract, and a book about theses should meet the same test. Fru 7

Page 10 of 214.

Page 11 of 214. Introduction The Turning Point We are going through a time of radical transformation. In under four years, software that can read, write and reason has gone from a curiosity to something every company is expected to have a plan for. Your customers are being asked to decide what their companies will look like when every employee has an AI working beside them. You are being asked to sell them things that did not exist a year ago. This book is for the people who do that selling: the sales engineer, the account executive, the technical consultant, anyone whose job is to walk a large enterprise to a decision. The hard thing about change is that you have to change. Buying is the easy part. Your customer can sign a contract in a quarter. Getting several thousand of their people to work differently is the transformation, and no product does that for them. A seller who only delivers the product has deliv‐ ered the easy part. So how does a large group change direction together when nobody can steer each member of it? 9

Page 12 of 214. Thesis-Based Selling Watch starlings at dusk. Several thousand birds wheel across the sky as one shape, turning, folding, opening out again. It is called a murmuration, and nobody is in charge of it. No bird gives orders, and none of them holds a plan of the whole. Researchers who filmed the flocks over Rome found that each bird tracks six or seven of its neighbors and follows a few simple rules. That is enough. A handful of rules, held by every bird, carries thousands of them through a turn that no leader could direct. Organizations change the same way. Nobody can instruct every person through a transformation. What turns them is a few simple things that everyone holds as true, from which each person can work out the next move. I call those things theses. A thesis is a belief about how the world works, stated plainly enough that everyone in the cus‐ tomer’s company, and everyone in the deal, can act on it. Bringing one is now the seller’s job. Before I make that case, one word needs pinning down, because the whole book leans on it and people use it to mean very different things. What I mean by AI When I say AI in this book, I mean one kind of AI. The term has been around since the 1950s, and most large companies have used some form of it for years. Machine learning models score loan applications and forecast demand. Decision support systems sit beside managers and suggest the next step. Those systems learn patterns from data or follow rules people wrote, and each does one job it was built for. They are predictable, and a buyer can test them the way they test any other software. This book is about something different: generative AI, the large language models released to the public from late 2022, and the systems built on them, led by labs such as OpenAI and Anthropic. These AI models do not follow fixed rules. They work on probability, producing the most likely next word, line of code or step of a plan, which is why the same question can get two different answers and why they can do jobs nobody built them for. That is what makes them powerful, and it is what makes them hard to buy. A buyer cannot test a probabilistic system once and know how it will behave next quarter, and the AI model under it will be replaced several times before the contract ends. 10

Page 13 of 214. The Turning Point That is the AI your buyer is asking about, and the AI this book means every time it says the word. With that settled, start with the person across the table, because some‐ thing about them has changed too. The consumerized C-suite One thing about selling AI has no precedent in enterprise software, and I think it is the most important thing to understand before anything else in this book. The person you sell to already uses it. Walk into an enterprise to sell an AI tool, and you can be sure of it. Your buyer uses ChatGPT. Your buyer uses Claude. Your buyer uses Gemini. Most likely on a personal account of their own, at home, for things that have nothing to do with your product. Technology people have a name for consumer technology that reaches work through the people rather than through the purchasing department. Around 2001 two researchers at the Leading Edge Forum, Douglas Neal and John Taylor, started calling it the consumerization of IT. For twenty years it mostly described the rank and file: the employee who brought a personal phone to work and expected the company’s email on it. AI has carried consumerization to the top floor. The executives who sign, the C-suite of chief executive, chief financial officer, chief operating officer and the rest, now use the technology themselves, at home, before anyone has bought it for them. I call that the consumerized C-suite, and it changes the meeting more than any feature on your roadmap. Enterprise software was never like that. Think of the systems sellers used to sell: an ERP, the enterprise resource planning system that runs finance and supply chains, from a vendor like SAP; a CRM, the customer relation‐ ship management system the sales team lives in; a data warehouse. The exec‐ utive who signed the check never logged into any of them. They had no picture of what good looked like, so they relied on others. Their team ran the proof of concept and validated it, and the executive signed to affirm it. This time the person across the table uses AI too, whether they are a vice president, a chief executive or a member of the board. Picture a chief oper‐ ating officer listening to a vendor describe how their assistant will summarize her contracts. While the seller talks, she opens the assistant on her own phone, pastes in a clause, and reads the answer. When the seller says the 11

Page 14 of 214. Thesis-Based Selling product is “highly accurate,” she already knows what accurate looks like and what it does not, because she checked it herself the night before. She has an AI model of her own now, and she asked it about you before the meeting. That changes the job. A consumerized C-suite has expectations, and it knows what good looks like. You cannot do smoke and mirrors and get away with it. Selling to this buyer takes a rigor that most sellers are not used to. Vague claims get tested in the room. A demo that only works on the prepared example gets a second, harder example from the buyer’s own phone. In my experience the seller who wins here is the one who says plainly what the product does well, where it still struggles, and why the buyer should go ahead anyway. That last part, the why, is where this book is headed. The question every buyer asks Every enterprise sales meeting arrives, sooner or later, at the same question. THE QUESTION Who else is doing this? Sometimes it comes dressed up. Can you share a customer in our industry? Is there someone we could talk to? What results are others seeing? It is one ques‐ tion, and I think it is the most reasonable question a buyer can ask. The buyer is about to spend real money, and some of their own reputation, on a 12

Page 15 of 214. The Turning Point decision they cannot fully check. So they ask to borrow somebody else’s certainty. Show me a company like mine that did this and came out fine. For decades, sellers had an answer ready. A slide of logos. A case study with a number in the headline. A reference call with a friendly customer. Geoffrey Moore described the logic in Crossing the Chasm in 1991: the prag‐ matists who make up the bulk of any market buy when other pragmatists have already bought, and they want references from people like themselves. Whole sales organizations were built to produce those references and put them in front of the next buyer. Now put the question to someone selling AI. THE QUESTION Who else is running this in production, at our scale, for long enough to show it worked? The honest answer is almost nobody. The technology buyers are asking about reached most companies in late 2022. The AI models have been replaced several times since. The prices have fallen by more than two orders of magnitude. And the early studies of what companies have gotten back from their generative AI spending are sobering; Chapter 3 looks at one of them. The few success stories that exist describe decisions made on tech‐ nology that has already been replaced. Picture the request. A buyer asks for a reference customer that has run AI agents in production for five years. A job post asks for five years of experi‐ ence building AI agents. Nobody can meet either one. The technology has been in most companies for about three years. Everyone is blazing the trail as they go. So the seller reaches for the oldest tool in the bag and finds the bag almost empty. Some sellers fill it anyway. They stretch a pilot into a success story, or quote a number from a customer whose situation has nothing in 13

Page 16 of 214. Thesis-Based Selling common with this one. Buyers can tell, for the reasons above: they use the technology themselves. I think there is a better answer, and it is the subject of this book. THE ANSWER Nobody has done this long enough to prove it. Here is what we believe is true about where this is going. Here is what would prove us wrong. Here is what follows for you if we are right. Tell me where you disagree. That answer does not point at anyone. It lays down a position and offers to defend it. The old word for a position laid down to be defended is a thesis. Think of a hike. On a marked trail you follow the trail map. Somebody has walked it before, the turns are drawn, and the signs tell you how far it is to the top. That is what references, case studies and analyst reports have been for an enterprise buyer: a map of a route others have walked. Now leave the trail. Where nobody has walked, the map has nothing to show you. Hikers who go off trail do not give up; they fall back on the fundamentals. They know where north is. They know the direction they are heading and why. They check the compass at every ridge, and they correct when the ground tells them they have drifted. No trail does not mean you cannot find your way. It means you navigate by something more basic than a map. A thesis is that compass. It tells a buyer which direction is right before anyone has walked the route, and it can be checked and corrected along the way. It is how decision makers have to decide now, and it is what this book teaches you to bring them. 14

Page 17 of 214. The Turning Point Why I am writing this For more than fifteen years I have been a sales engineer and a leader of sales engineers, selling data, AI and change to some of the largest and most complex enterprises in the world. I have watched buying patterns come and go, and I have seen deals won and lost in the trenches. I use AI every day. I am writing this book because I have watched a thesis change before, and I know what it does to sellers who do not see it. I sold the move to the cloud, and I saw very good products lose to a belief. Chapter 2 tells that story. AI is a bigger change of the same kind, and this time the reference customers have not had time to appear. Buyers are being asked to rethink what they believe about software, about data and about their own staff, with nobody to copy. They are looking for a thesis. Most sellers hand them a feature list. Almost everywhere else in business, the belief comes first. Investors write an investment thesis before they write a check. An investment thesis is a short, written view of where a market is going and why a bet on it will pay. Picture a fund that writes: every mid-sized company will run its accounting in the cloud within ten years, so we back the tools that make that move easy. Every deal the fund looks at gets measured against that sentence; Chapter 5 shows how one works in full. Builders do the same. SpaceX was built on the belief that rockets could land and fly again, and that reuse would make reaching orbit far cheaper. Apple built the iPhone on the belief that the phone would become a computer you carry in your pocket. In both cases the belief came first and the product followed, as Chapter 2 tells. Selling does it the other way round. We lead with the product in front of us and hope the belief takes care of itself. And the books we learn from do not help with this part. The best books for sellers are good at the product side. John Care and Aron Bohlig’s Mastering Technical Sales teaches the craft of the sales engineer. Robert Riefstahl’s Demonstrating to Win! teaches the demo. The Trusted Advisor, by David Maister, Charles Green and Robert Galford, teaches how to earn trust. All three assume that the buyer and the seller already share a picture of the world. That picture is what is changing, and that is the gap this book is written to fill. The Note at the front named the three shifts that open it: the consumerized C-suite, the procurement time trap and the missing reference. You have already met the first and the third in this Introduction, in the 15

Page 18 of 214. Thesis-Based Selling consumerized C-suite and in the question every buyer asks. The next section is about the speed underneath the second, and Chapter 2 gives the trap its full treatment. A paradigm shift in business software In 1962 the physicist and historian Thomas Kuhn published The Structure of Scientific Revolutions. He called the shared picture a field works inside a paradigm. Inside a paradigm, people argue about the details and agree on the ground. Every so often the ground itself gives way. He called that a paradigm shift. Business software is going through one. Until now, software was a tool that a person used. Now one employee can direct five or ten AI agents, each doing work that used to need a person. Work can be handed off at a scale most companies have not begun to picture. And the picture will not hold still. The leading AI models change in ways that matter every few months. METR, a research group that measures what AI can do on its own, found that the length of task an AI agent can finish has doubled about every seven months since 2019. Anything I write today about what the AI models can and cannot do will be out of date within a release or two. Enterprises do not move at that speed. Getting a new capability into a company’s workflows and systems takes years. So there is always a gap: more capability on offer than any company has absorbed. Your buyer lives in that gap. Every decision is now a bet Inside that gap, choosing AI is less like picking a tool and more like betting on the future. Business history is full of companies that bet on the wrong one. Consider three of them. Netflix began by mailing DVDs, then bet that people would watch over the internet. Blockbuster bet on its stores, and in 2000 it passed on a chance to buy Netflix for about $50 million. It filed for bankruptcy in 2010. Kodak’s story is subtler. A Kodak engineer, Steven Sasson, built the first digital camera in 1975, and Kodak later sold digital 16

Page 19 of 214. The Turning Point cameras in large numbers. But the company kept betting that film, paper and printing would stay at the center of photography, because that was where its profits were. When the phone became the camera, the bet failed, and Kodak filed for bankruptcy in 2012. Borders bet on its shops. In 2001 it handed its online bookstore to Amazon to run, ran its own website again only in 2008, and filed for bankruptcy in 2011. None of them lacked good products. Blockbuster’s stores were well run, Kodak’s film was excellent, Borders’ shops were pleasant places to buy a book. Each held the wrong thesis about the future, and no amount of product quality could make up for it. For a seller, those stories are worth keeping close, because they are what your buyer is afraid of. Much of the fear of AI is less the fear of the tech‐ nology than the fear of placing the wrong bet and becoming the next cautionary tale. A seller who understands that does not sell features to a frightened executive. They help the executive see the bet clearly, and decide it. And every decision maker now has to place one, on two horizons. The next six months are fairly easy to call: which tools to pilot, which teams to start with, what to measure. The next two to five years are not: what work will look like, what to build in house and what to buy, which vendors will still matter. A leader needs a view of both, and the two have to be consistent: the six-month plan should be a first step on the five-year bet, not a separate decision. A seller who can talk about both horizons, and connect them, is talking about what the leader actually lies awake over. If that sounds less like buying software and more like investing, it is. A company that commits to foundational AI is putting money, data and its people’s time behind a change that nobody has finished yet, the way a venture fund leads a company’s first large round on a belief about where it is going. Investors do not decide those bets with a feature scorecard. They write a memo that says what they believe and what would prove them wrong. Chapter 5 takes that parallel apart, because it tells you what kind of document your buyer actually needs from you. 17

Page 20 of 214. Thesis-Based Selling Choosing AI is no longer picking a tool. It is picking a thesis about the future. What selling is A bet is placed by a person, and a person has to be persuaded to place it. That brings us to the seller, and to a question that sounds too simple to ask. If you are reading this, you sell in some way, or someone who does asked you to read it. So it is fair to ask what selling is. The best definition I know came from one of my sales leaders, and over the years it became my own belief: sales is a transfer of energy. If you believe in what you are selling, selling gets much easier. You do not need a script. The conviction comes through in how you talk, and the other person feels it. People can sell things they do not believe in. Some do it every day. It is harder, and buyers can usually tell. Believing in what I sold has served me well in my career. “ Belief is the energy a buyer can feel. Thesis-Based Selling does not replace whatever selling means to you. It gives the energy a foundation. A thesis is what you believe, written down and shared. Once you hold one and believe it, the energy has something solid to carry. Your conviction stops being about the product and becomes a view of where the buyer’s world is going, and anyone in the room who comes to share that view starts carrying the energy with you. The Thesis-Based Selling framework Belief is the fuel. It is not yet a method. Ideas do not change deals on their own; moves you can repeat do. That is why sales teams adopt named methods. Selling is a hard job, and the people who last in it rely on frame‐ works because a framework gives a team one language, one set of steps to run 18

Page 21 of 214. The Turning Point on every deal, and something a manager can coach against. I did not want this book to be theory. A framework turns a belief into steps you can run on Monday, on a real account, with a real customer. The Thesis-Based Selling framework has one goal. It uncovers the few beliefs a customer’s decision rests on, gets the room to agree on them, and builds and defends the deal on top. Take one of your own live accounts through it, and you come out with two things: a thesis written down that the buyer agrees with, and a deal built on it. It runs in two rounds of three steps. ROUND ONE 1 Establish the thesis 2 LIST every reason ROUND TWO 1 on your own ASK WHY 3 down to a belief WRITE IT DOWN one sentence each Sell on the thesis AGREE your team, then the buyer with other people 2 BUILD 3 if this is true, then DEFEND what survives REVISE AND GO AGAIN Figure 1 Two rounds of three steps: every defense sends you back to a better thesis. Round one is Establish the thesis. It is discovery, and you can do it alone. 1 List. Write the decision and every reason behind it. 2 Ask why. Under each reason, ask why it matters until you reach a belief. 3 Write it down. One sentence for each belief, and what would prove it wrong. 19

Page 22 of 214. Thesis-Based Selling Round two is Sell on the thesis. It is the deal itself, and you do it with other people. 1 Agree. Get the people in the deal onto common ground. Your own team first, then the buyer. 2 Build. Work out what follows. If the thesis is true, an implication follows from it, a decision follows from the implication, and a product comes last. 3 Defend. Take the pushback, concede what is right, and keep what survives. Each round works on its own. Round one alone gives you a sharper view of any deal, even one you are not ready to take to the buyer. Round two alone helps when a thesis already exists and the work is getting people onto it. Together they run the whole deal. Notice the line in the figure that runs from Defend back to the start. Defending a thesis always teaches you some‐ thing: a reason you missed, a belief that was weaker than you thought. You take that back to round one, write a better thesis, and go again. The work‐ book at the back has a page to write on for each step. A word on what kind of framework this is. It comes from fifteen years of enterprise deals and the move to the cloud, not from a study of recorded calls. MEDDIC (named for Metrics, Economic buyer, Decision criteria, Decision process, Identify pain and Champion), Gap Selling and Solution Selling, which Chapter 4 explains, are practitioner frameworks of the same kind; where research books such as SPIN Selling, The Challenger Sale and The JOLT Effect have findings that bear on it, I cite them. And I hold the framework to the standard it asks of every thesis, so here is what would prove it wrong. The framework is wrong if buyers in AI deals turn out to decide on references after all, as references appear, and agreeing a thesis does not move them. It is wrong if teams that run thesis sessions, the working meet‐ ings where a team agrees its theses, and keep a ledger of how each thesis held up, see no change in no-decision rates or cycle length against their own base‐ line. And it is wrong if theses laid down in a first meeting close conversations more often than they open them. Test it on your own pipeline; the Field Guide at the back says how. 20

Page 23 of 214. The Turning Point Pick a real customer now, one deal you are working this quarter, and carry it through the book. Part One, Why Selling Has to Change. The ground the rest stands on: the path enterprise selling was built on, the belief under every large purchase, the reference customer that has gone missing, and what the sales methods you already know take for granted. Part Two, Establish the Thesis. Round one in practice: what makes a thesis, how to dig three to five out from under a decision, and the five I hold about AI, which serve as the running example from then on. Part Three, Sell on the Thesis. Round two in practice: agreeing on the theses with your own team and the buyer, building the sale on them, defending them, and using a language model as your opponent before a customer becomes one. The workbook closes the book. If you take one thing from this book, take this. Sell with conviction. Make a prediction about where your customer’s world is going, say it out loud, and be ready to defend it. A thesis is a bet you are willing to put your name to. The sellers who win the next few years will not be the ones with the longest feature list. They will be the ones who believed something specific, early, and said so. 21

Page 24 of 214.

Page 25 of 214. Part I Why Selling Has to Change Lay the foundation: how enterprise selling was built, why people buy the thesis, why the reference customer is gone, and why the frameworks we have no longer fit.

Page 26 of 214.

Page 27 of 214. Chapter One The Proven Path, and Why It Is Breaking “Past performance is no guarantee of future results.” The standard investment disclaimer IF THIS IS TRUE Enterprise software was sold for fifty years on a proven path, and the path worked because software held still. THEN Software that no longer holds still needs a new way to sell, and the old path cannot supply it. 25

Page 28 of 214. Thesis-Based Selling The Introduction made the case in outline: the reference customer has run out, your buyer already uses AI, and every decision has become a bet on the future. Before I offer a new way to sell, it is worth looking properly at the old one, because it was not a mistake. It worked for fifty years. If we understand what it rested on, we will know exactly what has gone missing, and what has to take its place. Enterprise selling is one of the most studied jobs in business. It has its own schools, its own methods and its own shelf of books. It has a process so well worn that a new account executive can be handed it on day one and run a deal by the end of the quarter. Discovery. A request for proposal. A demo. A proof of concept. Procurement. A signature. That process was not invented by accident. It was built, step by step, by people who studied what worked and wrote it down. For a long time it was the right way to sell, and following it was the smart thing to do. The process matters, and so does where it came from, because once you see what the path was built on, you can see why it is giving way under our feet. How selling became a discipline Start with a copier. In 1959 a small company in Rochester, New York, called Haloid Xerox, brought out the Xerox 914, the first plain-paper office copier. It was a big, expensive machine, and most offices could not buy one outright. So Xerox leased it: about $95 a month, with the first 2,000 copies included and a few cents for every copy after that. Customers paid for what they used. Hold on to that. It comes back at the end of this chapter. A machine sold that way needs a lot of salespeople, and Xerox took training them seriously. By the late 1960s it had a course called Professional Selling Skills. It taught a seller to find the customer’s need, present features as benefits, handle objections and close. Xerox was so sure of it that it set up a business, Xerox Learning Systems, to sell the course to other companies. In 1974 it opened a training center on more than two thousand acres in Lees‐ burg, Virginia, where new sellers lived and learned the method together. Xerox became the closest thing selling had to a university. Its sales force took part in an early pilot of the research Neil Rackham would later publish as SPIN Selling. One of its sellers, Michael Bosworth, studied what the best 26

Page 29 of 214. The Proven Path, and Why It Is Breaking people at Xerox Computer Services did, left in 1983, and founded the training company that became Solution Selling. Chapter 4 walks through these methods and the problem each one solved. The point here is simpler. By the 1980s, selling complex products to large companies had stopped being a talent and become a discipline. You could teach it. You could inspect it. You could follow it. The proven path, step by step Here is an illustration of that discipline at work, built from the kind of deal that was run thousands of times. A regional bank decides, in 2005, that it needs a new system to manage its customers. Nobody in the room doubts what such a system is. They have seen one. Their competitors have one. So the bank’s team does what it knows. It sends a request for informa‐ tion, an RFI, to a dozen vendors to see who is out there, and writing it, waiting for the answers and reading them takes six weeks or so. Then it writes a request for proposal, an RFP, with three hundred questions, most of them copied from the last one, because everyone already knows what a customer system has to do. The vendors take a month to answer and the committee takes another to score them. Five vendors answer. Three make the shortlist. Each of the three comes in for a scripted demo. The bank hands them the same scenarios and scores them on the same sheet. Then come the work‐ shops: two or three sessions with each vendor, for the people who will use the system, the architects who will run it and the finance team who will pay for it, and finding dates that suit all of them takes a month on its own. Two vendors get a proof of concept, a POC: a trial of a few weeks to a couple of months, with real data on a test server, to show the software does what the demo promised. One wins. Then the security team sends its questionnaire, about where the data will live and who can reach it, and the lawyers read the contract, and those reviews take several weeks more. Procurement takes over, the price comes down, and the contract is signed on the last day of the vendor’s quarter, because that is when the discount was biggest. Add it up and nine months from the first letter to the signature is normal, and a year is not unusual. Nobody in the bank thinks of that as slow. It is what careful looks like. 27

Page 30 of 214. Thesis-Based Selling Every step on that path takes risk out of the decision. The RFP says the product does what the category does. The demo shows it. The proof of concept proves it on your data. The reference call proves it worked for someone like you. By the time the executive signs, almost nothing is left to believe. It has all been checked. That is the beauty of the proven path. It turns a large, frightening bet into a series of small, safe checks. The license era: sold once For most of that history, the bank would have bought a license. A license was a right to run the software forever, paid for up front. The customer ran it on its own machines. On top of the license, the vendor charged maintenance every year, usually around a fifth of the license price, for support and upgrades. That business model shaped the seller. The big money came in at signature, so the seller’s work was mostly done at signature. Close it, book it, move to the next account. Somebody else installed it. Some‐ body else supported it. The seller came back when there was a new product to sell or a new department to sell it to. The license model also created its own temptation. In the late 1980s Oracle’s sellers pushed customers to buy as much software as possible up front, so the revenue could be booked in the current quarter. When the future sales behind those deals did not arrive, Oracle had to restate its earn‐ ings, and in 1990 it laid off about a tenth of its staff. The license model was not only a sales model. It was a very good busi‐ ness. In 2004 Siebel Systems, which had led the market for customer soft‐ ware, reported $487 million in new license revenue and $470 million in maintenance. The old customers were paying almost as much to keep the software as new customers were paying to buy it. The subscription era: sold again at renewal On February 22, 2000, Siebel held its user conference at the Moscone Center in San Francisco. Outside, a crowd of protesters marched with signs. Each 28

Page 31 of 214. The Proven Path, and Why It Is Breaking sign had the word software inside a red circle with a line through it. A camera crew filmed them for the evening news. The protesters were actors. The camera crew was fake. Both had been hired by a company barely a year old called Salesforce, whose founder, Marc Benioff, had a thesis: businesses should not install software at all. They should rent it over the internet and pay by the month. Siebel’s people called the police. The press called it a story, and it ran. What Salesforce sold was a subscription, priced per user, per month. The software ran on Salesforce’s machines, not the customer’s. That changed the seller’s job in a way that took the industry years to absorb. With a license, the customer paid once and was yours. With a subscription, the customer could leave at renewal. Salesforce learned that the hard way. In 2005, the year after its public offering, an executive named David Dempsey stood up at a company retreat and told the leadership that the company was in a death spiral. About 8 percent of its customers were leaving every month. They called him Dr. Doom. Benioff listened, and Salesforce built a team whose only job was to make sure existing customers got value from the product, so that they stayed. It is often told as the beginning of what the industry now calls customer success. So the motion grew a second act. Sell the seats. Hand the customer to someone who makes sure they use them. Come back at renewal and hope to sell more seats than last time. The consumption era: sold every day Then the meter came back. On March 14, 2006, Amazon launched a storage service, S3, and charged by the gigabyte. In August it launched EC2 and rented computers by the hour, starting at ten cents. You did not buy capacity in advance. You used it and got a bill. Cloud data platforms followed the same idea, and for them it became the whole business. Snowflake is the cleanest example. A customer commits to spend, but Snowflake earns revenue only when the customer actually runs queries. In 2022 its head of revenue operations wrote that consumption revenue “is earned through usage rather than generated from bookings and renewals,” 29

Page 32 of 214. Thesis-Based Selling and the company had already changed how it paid its sellers: a large part of their pay came from what customers consumed, not from what they signed. That is a different job. A contract is no longer a win. It is a promise the seller has to help the customer keep. The seller has to find the next workload, help it get built, and watch it run. And when customers use less, everyone feels it. On May 24, 2023, Snowflake told investors that some large customers were consuming less than expected, and cut its forecast for the year. The next day its shares fell by about 17 percent. Remember the copier. Consumption was not new. Xerox charged by the copy in 1959. What was new was that software, which had been sold like a building, was now sold like electricity. The newest step goes one further: pricing by outcome. In 2023 Intercom began charging for its AI support agent, Fin, by the customer question it resolved, 99 cents for each one, not by the seat or by the hour of computing. The seller is paid only when the result arrives. Outcome-based pricing makes the vendor carry more of the risk, and it changes the seller’s job again: the conversation is no longer about what the product can do but about what both sides agree will count as the result. That is already a conversation about what is true. FIELD NOTE 01 A license is sold once. A seat is sold again at renewal. Consumption is sold every day the customer uses it. Why the path worked Look at the three business models side by side. The license, the seat and the meter changed how the seller was paid and when the seller had to show up. They did not change the path. Every one of them still ran discovery, the RFP, the demo, the proof of concept and procurement. The path held because three things held under it. Software was software. A customer system in 2005 did what a cus‐ tomer system did in 1999, with better screens. The category was settled. The buyer knew what to ask for, and the RFP could be copied from the last one. 30

Page 33 of 214. The Proven Path, and Why It Is Breaking The signposts were stable. Because categories lasted, analysts could map them, references could pile up, and feature comparisons meant some‐ thing. A reference from two years earlier described a product that still behaved the same way. The product held still long enough to be proven. A proof of concept run in March was still true in September. A three-year contract bought roughly the same thing in year three that it bought in year one. On that ground, the winning formula was not complicated. Be good. Be likeable. Know the product. Tell a strong story. Win the feature comparison. Put your best reference customers in front of the buyer. If your product was good, that was usually enough. There was very little reason to reinvent how selling was done, because the thing being sold did not reinvent itself. “ Selling stayed the same for fifty years because the software held still. The proven path was proven on products that held still. Why the path is breaking Now listen to the people who build the software. In December 2024, Microsoft’s chief executive, Satya Nadella, sat down with the investors Brad Gerstner and Bill Gurley on their BG2 podcast. He described business applications as databases with “a bunch of business logic” on top, and said “the business logic is all going to these agents.” The notion that business applications exist, he said, is probably where they will collapse in the agent era. That was the head of one of the largest business software companies in the world, saying the shape of business software was about to change. Two months earlier, Google’s chief executive, Sundar Pichai, had told investors that more than a quarter of all new code at Google was generated 31

Page 34 of 214. Thesis-Based Selling by AI, then reviewed and accepted by engineers. At one of the largest soft‐ ware companies in the world, the machine was writing and people were checking. In February 2025 Andrej Karpathy, one of the founding members of OpenAI, gave the new way of building a name. He called it vibe coding: you describe what you want, accept what the AI writes, and “forget that the code even exists.” By November, Collins Dictionary had made it its word of the year. And in early February 2026, when an AI lab released add-ons that let its agent do legal, sales and marketing work, investors sold software stocks so hard that traders gave the week a name: the SaaSpocalypse, after SaaS, soft‐ ware as a service, the industry’s name for software sold by subscription. Shares in established legal and data software companies fell by double digits in days. Put those together and you get something the proven path was never built for. Software is being written faster than it can be adopted. The version a customer is rolling out is already behind the version the vendor has shipped, and the vendor’s version is behind what an AI model could write next month. Anyone can describe software into existence. The buyer’s own team can build a working tool in an afternoon by talking to an AI model. The question in the room is no longer only “which vendor?” It is “why buy at all?” The unit being sold is in doubt. Seats assume a person sits in each one. If agents do the work, how many seats does a company need? Even Sales‐ force, the company that taught the world to sell software by the seat, priced its first agents in 2024 at $2 per conversation. When people say software is dead, they are exaggerating. Companies still need systems that hold their records and enforce their rules. But the belief the proven path rested on, that software is a stable thing you can specify, demonstrate, prove and then buy for three years, is not true anymore. AI is still sold the old way Here is the strange part. Most enterprise AI is still sold on the proven path. 32

Page 35 of 214. The Proven Path, and Why It Is Breaking The seller sends a deck. The buyer sends an RFP. The questions in it describe what the product could do last quarter. A demo is scheduled. A proof of concept is run, and by the time it ends a new AI model has shipped. Then a contract is signed, often priced by the seat, for a product whose capa‐ bility will change several times before the first renewal. Meanwhile the buyer has moved. In a 2024 Gartner survey of more than six hundred business buyers, 61 percent said they would prefer to buy without a sales rep at all. And as the Introduction showed, the executive across the table uses AI on their own account. They do not need a seller to show them what the technology can do. They need help deciding what to believe about it. So we have a new product, a new buyer and an old motion. Nobody stopped to reset what selling should mean when the thing being sold will not hold still. THE QUESTION If the product will change before the contract ends, what exactly is the buyer agreeing to? What this means for the seller There is a fear under all of this, and sellers feel it too. If AI can write the soft‐ ware, can it also sell it? Look at what much of the job has been. A seller, or the technical consul‐ tant beside them, learns the product: the features, the release dates, the roadmap, how it compares. Then they carry that knowledge to the customer. Call it knowledge arbitrage. The seller knows something the buyer does not, and the meeting moves it across the table. AI is absorbing that work. A demo can be generated on one cus‐ tomer’s data. Presentations and scenarios can be tailored in minutes. Documentation and product knowledge are moving into the product 33

Page 36 of 214. Thesis-Based Selling itself, where a customer can ask and get an answer without a seller in the middle. That is part of why so many buyers, as the Gartner survey found, would rather not talk to a rep at all. So the seller’s value has to move up. Away from the feature-and-function trench fight, and toward the work no document can do: helping a room full of people work out what they believe about the future, and agree on it. That is less about technology and more about psychology, and good sellers already have the skill. They have been reading rooms their whole careers. The job is not going away. It is changing. The seller who relays facts is being replaced by the product. The seller who helps a company decide what is true about its future is worth more than ever. The rest of this book is about becoming that seller. Nobody knows where this lands I want to be honest about one thing. I do not know where this ends. Nobody does. Maybe agents absorb most of what business applications do today. Maybe the systems that hold the records become more valuable, not less. Maybe seats survive, or maybe every price becomes a meter. The people with the most money riding on the answer disagree with each other in public every week. I am writing this book anyway, and that uncertainty is the reason. When the map is accurate, you follow the map. That was the proven path, and for fifty years it served sellers and buyers well. When the map is out of date before it is printed, you need a way to decide without one: a short list of beliefs you can state, test with the buyer and defend. The rest of this book calls that a thesis. 34

Page 37 of 214. The Proven Path, and Why It Is Breaking The proven path told you where to walk. Now you have to know which way is north. The next chapter is about that north: the belief under every large purchase, and what happens to a sale when the buyer holds a different one from yours. YO U R TAS KS What to do this week Draw your path. Write down the steps your last three deals went through, from first meeting to signature. Mark the steps that only exist because the product was assumed to hold still. Date your proof. For your last proof of concept, write down the AI model and version it ran on. Check whether that is still what the customer would get today. Name your unit. Write down what your customer pays for: a license, a seat or consumption. Then write what happens to that number if agents do half the work. Ask the bigger question. In your next meeting, ask the buyer: What do you believe your company will build itself, and what will it still buy? Write down the answer. It is the beginning of their thesis. 35

Page 38 of 214.

Page 39 of 214. Chapter Two People Buy the Thesis Before the Product “People don’t buy what you do; they buy why you do it.” Simon Sinek IF THIS IS TRUE On a large decision, people buy the thesis before they buy the product. THEN The seller’s first job is to establish the thesis, and features only get compared among products that share it. 37

Page 40 of 214. Thesis-Based Selling Chapter 1 walked the path enterprise selling was built on, and showed why it held for as long as software held still. If that were the whole story, the answer would be to wait for the market to settle and the path to come back. It is not the whole story. Under every large purchase sits something the old path never needed to name: the belief the buyer is really buying. This chapter names it. By the end you will know what a thesis is, be able to spot one under any decision, and see why the seller who brings the right one has often won before a single feature is compared. Take a paper map. A good road atlas was a fine product, and you could make it better in all the ways products get better. More detail. Sturdier paper. A new edition every year. Every atlas rested on the same belief: the map supplies the terri‐ tory, and the driver does the finding. If you had sold atlases against a rival’s atlas, you would have sold on exactly those features, and the better atlas would have won. Then satellite navigation reached ordinary drivers. In May 2000 the United States stopped degrading the civilian GPS signal, and within a few years a small box on the dashboard could tell you where you were and what turn to take next. The first units were worse than an atlas on almost every feature a map seller would list. The screen was tiny. The maps inside had holes. The box needed power, and it cost many times what an atlas did. None of that mattered, because the box was built on a different belief about what a map is for. The atlas says: here is the territory, find yourself in it. The box says: I know where you are, so I will tell you the way. Once a driver accepted the second belief, the atlas was out of the comparison. Nobody weighed paper quality against a voice that said turn left. The atlas makers did not lose on features. They lost on the belief underneath, and no new edition could win that back. 38

Page 41 of 214. People Buy the Thesis Before the Product ATLAS AGAINST ATLAS BOX AGAINST BOX More detail Screen size Sturdier paper Map coverage A new edition every year never compared Price THESIS THESIS Here is the territory. Find yourself in it. I know where you are. I will tell you the way. Figure 2 Features are compared inside a thesis, never across two. Read the figure across. On the left, atlas against atlas: more detail, sturdier paper, a new edition every year, all resting on one thesis, here is the territory, find yourself in it. On the right, box against box: screen size, map coverage, price, resting on another, I know where you are, I will tell you the way. Between the two columns, nothing is compared. A driver never weighed paper quality against a voice that said turn left, because the two products were answers to different beliefs. That is the shape of every thesis change in this chapter: the features stay inside their own column, and the buyer chooses the column first. I use the word thesis for that kind of belief: a claim about how the world works that a product is built on. What the driver bought was a new thesis, and the product came with it. What I mean by a thesis The word is Greek. Thesis means a placing, a setting down, from a verb that means to put or to place. A thesis is a belief you set down where other people can see it. That matters more than it sounds. A belief that stays in someone’s head cannot be agreed with, argued with or built on. Once it is set down, a room can look at it together. There are other words for the same thing. A belief. A worldview. A premise. A conviction. In the Introduction I used Kuhn’s word, paradigm, which is a thesis that a whole field holds at once. Each of those words is close, 39

Page 42 of 214. Thesis-Based Selling and none of them is wrong. I chose thesis because it carries the idea of putting something down on the table, in front of other people, as the place to start. I could have used axiom, and I nearly did. In mathematics an axiom is the truth everything else is built on, and nobody argues with it. That is too hard for a sales conversation. A thesis is softer: it is common ground two sides agree to build on, and it can be tested and revised. Chapter 5 tells the rest of why I did not choose axiom. So in this book, a thesis is the foundational belief that a decision rests on: set down, agreed on and built on. For a seller, that definition is practical, not academic. If you can name the belief your buyer’s decision rests on, you know what the deal is really about. If you cannot, you are answering ques‐ tions without knowing what they are for. Everyone already lives inside a thesis Everyone lives inside a thesis, and most of the time nobody says it out loud. Look at four people and organizations and you can see it working. A family that believes a house is the safest place to keep its money will stretch for a bigger mortgage and keep little in the stock market. Show that family a better investment and they will not hear it, because the comparison they care about is between houses, not between a house and a fund. A company that believes it should own and run its own machines will compare one data center vendor with another for years, and never seriously price the cloud. That was the belief the cloud had to overturn, as the next section shows. A whole industry believed, until 2007, that a serious phone needs a keyboard. Every product review compared keyboards. The phone that won the next decade had none. And a buyer who believes their data will outlive any tool they run on it will rank open formats above speed, whoever is selling. In each case the thesis decides what gets compared and what never comes up. Learn to hear the belief under a decision, and you start to find theses everywhere. The rest of this chapter is practice. 40

Page 43 of 214. People Buy the Thesis Before the Product A thesis change I sold through: the cloud I used to sell the move to the cloud. At the time there was a lot of excellent software built to run in a company’s own data center. Data warehouse appliances like Netezza, and Hadoop clusters from vendors like Cloudera, could do remarkable things on machines you owned. On a feature sheet those products often won. They were fast, they were mature, and their makers kept adding ways to run them next to the cloud or partly inside it. Here is an illustration, built from the kind of deal I worked on in those years. A large manufacturer runs its reporting on an appliance in its own data center. It is a capable machine. It runs the queries and returns the answers. Its one weakness is that it will not grow. When the box fills up, the only move is to buy a bigger box, and that takes a budget cycle. The team choosing a replacement does what such teams do. They build a benchmark: the same forty queries, run on the appliance and on a data ware‐ house built natively in the cloud. The appliance holds its own. The incum‐ bent’s seller points at the numbers and says, reasonably, that you can run your queries here and you can run them there, so why move? That was never the question. The sales engineer on the cloud side puts the benchmark down and asks a different one. THE QUESTION “Where will your data come from in five years?” The chief information officer thinks about it. Machines on the plant floor. Suppliers. Distributors. Almost none of it will start life inside her data center. “And where do you believe computing is going?” She does not need long. “To the cloud. All of it, eventually.” 41

Page 44 of 214. Thesis-Based Selling “Then you can keep running on the appliance for a while. The question is how long before you are the last one there.” The benchmark never came up again. Nothing about the appliance had changed. She had said out loud what she believed about the future, and after that she was no longer choosing between two ways to run a query. She was choosing between a future she believed in and one she did not. The same thing happened across the industry. The on-premises products were built on the belief that a company owns and runs its machines. Buyers came to believe something else: the future will increasingly be in the cloud, and whether or not you can keep running on your own hardware, it is only a matter of time before you are left behind. We have since watched that come true. I see the same thing now with open table formats: ways of storing a company’s tables so that any database engine can read them, not only the one that wrote them. Apache Iceberg is the best known. People argue about whether Iceberg is faster or slower than some other way of storing a table. That argument misses why buyers ask for it. Iceberg carries a thesis: my data will sit in an open format that any engine can read, and my platform will be open. Interoperability is the belief. A buyer who holds it will pick an archi‐ tecture that honors it, and a buyer who does not hold it will never be moved by a benchmark. A benchmark only means something inside a thesis. Put a horse and an SUV through a speed test. The SUV wins, and it tells you nothing. Both will get you from one town to the next, and nobody chooses between them on speed. Two tools built on different theses are the same. Benchmark one against the other and you learn only which is faster at a race neither buyer is running. Benchmarks settle a question only after the thesis is agreed. The theses behind big technology decisions Look for it and you find it under most of the big decisions in technology. 42

Page 45 of 214. People Buy the Thesis Before the Product The internet. On May 26, 1995, Bill Gates sent his executives a memo titled “The Internet Tidal Wave.” Microsoft was then the most successful software company in the world, and its products were built on the belief that software lived on a personal computer. The memo replaced that belief. “Now I assign the Internet the highest level of importance,” he wrote, and he called it the most important development since the IBM PC. The memo announced no product. It changed the thesis, and the product plans were rewritten to follow. Picture a seller who walked into Microsoft that summer with a product built only for the desktop. The pitch could be flawless, and it would still be pitching last year’s belief. The iPhone. When Apple showed the iPhone in January 2007, Microsoft’s chief executive, Steve Ballmer, judged it by its features. It was “the most expensive phone in the world,” he said, and it “doesn’t appeal to business customers because it doesn’t have a keyboard.” On the feature sheet of 2007 he was right. A BlackBerry was the better email machine. Apple was selling a different belief: the phone is a com‐ puter, and its whole face should be software. Buyers who accepted that stopped counting keys. Android phones, which arrived the next year, do most of what an iPhone does: the same apps, the same cameras, often the same price. The thesis is different. Android is an open platform that many makers build on, so the buyer gets choice, many prices and a phone that can be changed. Apple is one company controlling the hardware and the software together, so the buyer gets one experience that works the same way everywhere. Neither is better on a feature sheet, and the feature sheets have looked alike for years. People who choose one over the other are choosing a thesis about what a phone should be, and they rarely switch. The same split runs through the history of the computer itself. Windows was built on a proprietary thesis: one company owns the code, sells licenses and stands behind the product. Linux, released by Linus Torvalds in 1991, was built on the opposite one: the code is open, anyone can read and change it, and a community improves it. Both run the world’s computers. A company choosing between them was never only comparing features. It was choosing who it wanted to depend on. The rocket. SpaceX was founded in 2002 on one sentence: humanity should become a multi-planetary species. Look at what follows from it. If people are going to live on another planet, the cost of reaching orbit has to 43

Page 46 of 214. Thesis-Based Selling fall by orders of magnitude. For that, a rocket has to be reused the way an airplane is. So the company spent years learning to land a booster, did it in December 2015, and flew astronauts to the space station in 2020. Each of those was a product decision, and all of them were predicated on the thesis. A supplier who came to SpaceX offering cheaper parts for a rocket used once and thrown away was answering a question the com‐ pany had stopped asking. The electric car. In 2006 Elon Musk set down Tesla’s plan on the company’s blog: build an expensive sports car, use the money to build a cheaper car, then a cheaper one still, on the way to sustainable transport. On the feature sheet of the time, an early Tesla lost to a petrol car on price, range and the time it took to refuel. People bought it anyway. They were not buying a way to get from one place to another; every car does that. They were buying a view of where cars, energy and software are going: that cars would be electric, that the software in them would keep improving after the sale, and that they wanted to be early. A seller of petrol cars who argued range against that buyer was arguing about the wrong thing. The interface. Around 2002 Jeff Bezos told Amazon’s teams that every team would expose its data and functions through service interfaces, and that teams would talk to each other in no other way. No memo was published. Steve Yegge, an engineer who was there, recounted the order in 2011. Once that thesis held inside Amazon, a seller whose system could not be reached through an interface was selling against the customer’s thesis. It did not matter how good the seller was. If your product runs against the thesis your customer holds, selling harder will not help. Have the thesis conversation first. 44

Page 47 of 214. People Buy the Thesis Before the Product The outsourcing thesis, and why it is being decided again One of the largest theses in enterprise technology was never about a product at all. It was about who should build things. From the 1990s through the 2000s, many large companies came to hold a clear belief: engineering and IT are not our core business, so we should not do them ourselves. They handed the work to system integrators, firms that design, build and run technology for other companies, and increasingly to global system integrators, the largest of them, with delivery centers around the world: Accenture, IBM, and Indian firms such as Tata Consultancy Services, Infosys and Wipro. The backdrop is well documented. The work of fixing date code before the year 2000 sent a wave of projects offshore, much of it to India, and showed a generation of executives that large engineering work could be done far away, at lower cost, by a partner. Offshore delivery became the default for many companies. Other companies held the opposite thesis: software is the business, so the engineers stay in house. Marc Andreessen’s essay in 2011, that software is eating the world, gave that side its slogan, but plenty of companies had held the belief long before. Neither thesis is right in general. Each is a bet about what a company should own. What matters for a seller is that the same product lands completely differently in each world. Pitch an AI tool to a company whose thesis is “we do not build in house,” and the tool has to fit how its partners work: their delivery model, their contracts, the people who will actually run it. Pitch the same tool to a company that keeps its engineers close, and it has to make those engineers stronger, or it will be seen as a threat to them. A seller who does not know which thesis the buyer holds will give the right pitch to the wrong company. And the question is open again. AI agents change the arithmetic of building. A company that outsourced its engineering because people were expensive and scarce can now ask whether a small team directing many agents could build in house what it used to buy through a partner. A company that kept its engineers may ask the reverse. Buyers are re-deciding their outsourcing theses right now, often without saying so. Ask which way your buyer is leaning before you pitch. The answer will tell you who your real competitor is: another vendor, the buyer’s integrator, or the buyer’s own team with a set of agents. 45

Page 48 of 214. Thesis-Based Selling Evidence loses to a thesis In 1847 Ignaz Semmelweis, a doctor at Vienna General Hospital, was trying to work out why so many women in his maternity ward died of fever after giving birth. He had the doctors and students wash their hands in a chlorine solution before they examined patients. Deaths in the ward fell from about one in ten to between one and two in a hundred. On any feature comparison he had won. He had the numbers, a cheap fix, and a result anyone could repeat. The profession rejected him anyway. Doctors of the time believed disease came from bad air and from imbalances inside the patient, and under that belief the idea that a physician’s own hands carried death made no sense. His appointment was not renewed. Handwashing became standard a generation later, after Pasteur and Lister had replaced the belief underneath. His evidence was sound. What he never changed was the thesis of the people he needed to convince, and until that moved, the evidence had nowhere to land. Every seller who has lost a deal while holding the better benchmark knows the feeling. Why sellers sell features instead Most of us who sell make the same mistake from the other side. We sell the product in front of us. We walk the buyer through features and functions, compare ourselves with the competitor line by line, and treat the belief underneath as someone else’s department. Sales engineers do it most of all, and I say that as one. We are hired because we know the product, so we show the product. On a small purchase that works. If you are buying a monitor, compare the specifications. On a large one, and above all on a change that will reorganize how a company works, the buyer is deciding what to believe. Features come second. The buyer is choosing a picture of the future and then looking for a vendor who fits inside it. If your competitor supplied the picture, you are being scored against it, and you lose points for every way your product is different. 46

Page 49 of 214. People Buy the Thesis Before the Product FIELD NOTE 02 Features are compared inside a thesis. Whoever sets the thesis decides what gets compared. This is why a strong thesis outweighs a long feature list. A strong thesis makes some features irrelevant and others decisive. The cloud thesis made on-premises speed irrelevant and elasticity decisive, and it did so before any product was demonstrated. “ The strength of the thesis matters more than the length of the feature list. It is also why AI has put so many buyers in an uncomfortable spot. A company that spent ten years acting on one set of beliefs about software, data and staffing is being asked to question all of them at once. Its leaders are looking for a thesis they can hold. A seller who arrives with one, and can defend it, brings the thing they are short of. Speed on your own machines Elasticity IRRELEVANT DECISIVE THESIS The future is in the cloud Figure 3 The thesis decides which features count, before any product is shown. Read the figure from the bottom. One belief sits at the base: the future is in the cloud. Above it, two features that buyers compared for years. Speed on your own machines was the appliance seller’s strongest argument, and once the thesis was agreed it became irrelevant. Elasticity, the ability to grow and shrink with the workload, became decisive. Nothing about either feature changed. The thesis underneath changed, and it re-sorted the list before any product was shown. 47

Page 50 of 214. Thesis-Based Selling That is the pattern to look for in your own deals. When a buyer ranks features, the ranking is coming from somewhere. Find the thesis that produces it, and you know which of your features will count before the eval‐ uation starts. The procurement time trap: why an RFP cannot specify AI Chapter 1 walked the proven path: the RFI, the RFP, the demo, the proof of concept and a contract to renewal. It works when the capabilities are known and stable. Now put a clock on it, using the bank’s deal from Chapter 1. In my experience the path at a large company runs something like this. Six weeks or so for the request for information, from writing it to reading the answers. Two months or more for the request for proposal: writing it, waiting for the vendors, scoring what comes back. A month of vendor demos and work‐ shops with the teams who will use the product. A POC of six weeks to a couple of months. Then several weeks of security and legal review, often side by side, while procurement negotiates the price, and only then the decision. Nine months from the first letter to the signature is normal; a year is common. In those nine months the leading AI models have leapfrogged each other more than once, and the product the committee tested in its POC is not the product it will be running by the first renewal. The process was built to make a careful, durable decision, and it does exactly what it was built to do. That is the trap. How do you make a durable decision about a tech‐ nology that changes faster than your buying process? I call this the procure‐ ment time trap, and it is the second of the three shifts the Note at the front named. That machine is how large enterprises buy in the United States and in markets like it. Public sector buying runs on formal bids and contracts with rules of their own, and other countries and cultures buy differently again, but the shape will be familiar to most readers of this book. The trap is not only the calendar. It is also the document at the heart of the process. Try writing the RFP for an AI agent platform. Which capabili‐ ties do you require, and at what level? Picture a requirement a buyer might reasonably have written in early 2024: the system must summarize documents 48

Page 51 of 214. People Buy the Thesis Before the Product of up to fifty pages. Within a year the leading AI models read documents hundreds of pages long, and the line that took a committee a week to agree describes nothing anyone would buy. That is what happens to the whole RFP. By the time the responses come back, a new AI model has shipped. Some of your requirements are stale. Others nobody thought to write down. And the three-year contract has to outlive several generations of the thing it buys. Buyers and vendors had already started to bend the machine before AI arrived. Consumption pricing let a buyer pay for what they used instead of guessing at capacity up front. Outcome-based pricing, described in Chapter 1, ties the price to a result both sides agree on. Both are ways of buying when the future is hard to specify. AI pushes the problem past what pricing alone can fix, because the thing that cannot be specified is not the volume. It is the capability itself. The job now: pick the thesis and bet on it When the requirements cannot be written down, what is left is the thesis. The buyer has to decide what they believe and build the plan on it. So does the seller. In practice that changes what a buyer asks of a seller. Instead of “does your product meet requirement 4.2?”, the useful question becomes “what do you believe will be true in two years, and how does your product fit if you are right?” Instead of scoring answers to three hundred questions, the buyer agrees on three to five beliefs and asks which vendors fit them. The seller’s job moves with it: from answering the RFP to helping write the beliefs it should have been built on. Written down, those beliefs become a different kind of buying docu‐ ment. An RFP checklist asks whether a product has a list of features. A written thesis tells the buying committee what bet it is placing, why, and under what conditions the bet succeeds or fails. The committee can govern a purchase with that: it knows what to measure, when to look, and what would make it stop. Chapter 9 shows one being written, in a deal, from start to finish. That is a bet, and the stakes are the ones the Introduction described. Pick the right thesis and you reap it for years. Pick the wrong one and it can end a 49

Page 52 of 214. Thesis-Based Selling career or a business. A seller who helps a buyer make that bet well is remem‐ bered long after the product has been replaced. The job now is to pick the thesis, bet on it, and build the sale around it. YO U R TAS KS What to do this week Find the thesis under your product. Finish the sentence: This product only makes sense if you believe that… Find the thesis under your competitor’s. Do it fairly. If it is the same thesis, you are in a feature fight. If it is different, you are in a thesis fight and should stop arguing features. Recall one deal you lost while winning on features. Write down the belief the buyer held that made your features beside the point. Listen for it. In your next customer meeting, write down every sentence that starts with “we believe,” “we have decided,” or “our direction is.” Those are the buyer’s theses. 50

Page 53 of 214. Chapter Three Selling When There Is No Reference Customer “I think it’s important to reason from first principles rather than by analogy.” Elon Musk IF THIS IS TRUE In AI, the reference customer does not exist yet. THEN Selling by pointing at someone else’s success fails, and the seller has to reason with the buyer instead. 51

Page 54 of 214. Thesis-Based Selling In the last chapter we saw that people buy the thesis before they buy the product. If we stopped there, you would have a good idea and no reason to change how you sell tomorrow, because for most of the history of selling, sellers could get by without noticing the thesis at all. They had another tool that did much of the work for them: somebody else’s success. In this chapter I want to look hard at that tool, the reference customer, and at the analyst reports, benchmarks and best practices that grew up around it. I will show why, in enterprise AI, the tool has gone missing, and why reasoning from first principles, on its own, will not replace it. By the end you will see why the seller’s real job has become getting a room to agree on what is true. There are two ways to know that something will work. The first is that somebody already did it. You look at their result and expect roughly the same for yourself. This is reasoning by analogy, induction from precedent: it worked there, so it will work here. It is how most deci‐ sions get made, because it is fast and usually right. It has one condition. There have to be cases like yours to reason from. The second is to work it out from what has to be true. You strip the question down to the parts nobody can argue with, and build back up from them. This is reasoning from first principles, and its shape is deductive: from premises everyone accepts to a conclusion that follows from them. Aristotle called such a starting point a first principle: the first basis from which a thing is known. To see what the second way looks like in practice, take the example that made it famous in business. It is a battery. In the early days of Tesla, common wisdom said battery packs cost about $600 per kilowatt-hour and would stay expensive, because they always had been. Elon Musk asked a different question. What is a battery made of, and what do those materials cost on a metals exchange? Cobalt, nickel, aluminum, carbon, some polymers, a steel can. Bought as raw materials, he said, they came to about $80 per kilowatt-hour. The $600 was a habit and the $80 was a fact, and a company could be built in the gap between them. Notice what Musk did not do. He did not go looking for a carmaker that had already made batteries cheap, because there was none to copy. He reasoned from what had to be true. That is the position an enterprise AI 52

Page 55 of 214. Selling When There Is No Reference Customer seller is in more often than not: the buyer asks for a precedent, and the honest answer is that the precedent is still being made. First-principles thinking is good advice, repeated so often that it has turned into a poster. I want to look at what it asks of someone who sells, and I will come back to it at the end of this chapter, after looking at why the first way, reasoning by analogy, has stopped working. Selling runs on analogy Look at the tools a seller is handed on the first day. The customer story. The reference call. The slide of logos. The analyst’s Magic Quadrant, which this chapter explains below. Every one says the same thing: someone like you already did this, and it went well. Buyers have good reasons to want it. A buyer cannot test every claim a vendor makes, so the buyer borrows the result of someone who already ran the test. Robert Cialdini named the instinct social proof in 1984. Geoffrey Moore built a theory of technology markets on it in 1991, in Crossing the Chasm: the mainstream buyers he called pragmatists will not buy until they can see other pragmatists, companies like theirs, already succeeding. An older generation of buyers had a proverb for it: nobody ever got fired for buying IBM. FIELD NOTE 03 A reference is someone else’s result, borrowed. A buyer who cannot test your claim asks who already did. For most of the history of enterprise software the system worked, because the product changed slowly enough for references to pile up. A database sold in 2010 behaved in 2012 the way the 2010 case study said it would. 53

Page 56 of 214. Thesis-Based Selling The missing reference: three things broke at once The Introduction asked the question every buyer asks, who else is doing this, and the Note at the front called the empty answer the missing reference, the third of the three shifts. Here is why the answer is empty. Three things broke at once. The technology is young. Generative AI reached most companies after November 2022. A real reference takes time to grow. Someone has to buy, build, go to production, run it for a few quarters and measure. There have not been many quarters. The ground keeps moving under the stories that do exist. Stanford’s AI Index reported that the price of running an AI model at the level of GPT3.5 fell from $20 per million tokens (a token is the unit AI models read and write in, roughly three quarters of a word) in November 2022 to 7 cents in October 2024, a drop of more than 280 times in under two years. Capability moved as fast as price. A case study written eighteen months ago describes a decision made at a cost, and with an AI model, that nobody would choose today. The customer in the story would decide differently now. Most of the stories are pilots. In 2025 MIT’s NANDA project inter‐ viewed 150 leaders, surveyed 350 employees and studied 300 public deploy‐ ments. Its report, The GenAI Divide, stated that 95 percent of organizations were getting no measurable return from generative AI. The study’s methods and definitions have been widely questioned, so treat the number as a signal rather than a measurement, but the direction matches what most sellers see: many pilots, few in production. A pilot that never reached production tells the next buyer nothing, however good the logo looks. So when the buyer asks who else is doing this, the seller has three bad options: a story that is too young, a story that is out of date, or a pilot wearing a production badge. 54

Page 57 of 214. Selling When There Is No Reference Customer WHO ELSE IS DOING THIS? TOO YOUNG OUT OF DATE A PILOT not enough quarters old price, old AI model never in production Figure 4 Three answers to the buyer’s question, and none of them is a reference. The figure puts the three bad options side by side, and each fails for a different reason. The story that is too young has not had enough quarters to prove anything. The story that is out of date describes a decision made at an old price, on an old AI model. The pilot never reached production, so it proves only that someone was willing to try. A buyer who has read the MIT study can tell the three apart. So the seller who leads with any of them spends credibility instead of earning it, which is why the rest of this chapter looks for a better answer. Why customer logos mislead There is a fourth problem, and it is the hardest to see. During the Second World War, American bombers flew over Europe, and many did not come back. Armor could save them, but armor is heavy. Put it everywhere and a bomber cannot fly as far or carry as much. So the question was where to put it. The planes that returned seemed to offer the answer. They came home full of holes, and holes can be counted. As the story is usually told, the damage was mapped across the aircraft, and it clustered on the wings, along the fuselage and around the tail. The intuitive answer was plain. Reinforce the places that get hit. The question reached the Statistical Research Group, a team of mathe‐ maticians working for the military at Columbia University. One of them was Abraham Wald, and he saw what the map left out. Every plane in the count had come home. A plane hit in the engines mostly did not, so its damage was never counted. The clean areas on the returning planes were not the safe areas. They were the places where a hit brought a plane down. 55

Page 58 of 214. Thesis-Based Selling So the armor belonged where the returning planes had no holes. The evidence everyone was looking at was real. It had been filtered by survival, and read as if it had not been. The lesson has a name now: survivorship bias. Now look at the slide every enterprise vendor shows. A logo slide is a page of customer logos, put up as proof: these companies bought from us, so you can too. It is a map of the planes that came back. Every customer on it survived. The slide cannot show the pilots that stalled, the projects that were quietly shut down, or the proofs of concept that never reached production. Those are exactly where a buyer would learn where the danger is. For a buyer, that means the slide says little about whether the product will work for them. For a seller, it means that leaning on survivors invites the wrong conclusion, and a sharp buyer knows it. Picture the moment. You put up twelve logos, and the head of architecture asks how many of those twelve are running in production today. If the honest answer is three, the other nine logos have just cost you more than they earned. In a mature market this matters less. There are enough survivors, over enough years, that the failures show up in the averages. A buyer choosing a payroll system can talk to dozens of companies, hear the bad stories along with the good, and learn from the whole pattern. In a young market like enterprise AI, you are looking at a handful of planes and guessing about the rest. The failures have not been counted yet, and many of them never will be, because nobody publishes a case study about the pilot they shut down. When the survivors cannot tell you enough, you reason the way Wald did, from what must be true. In a sale, that means setting down what has to hold for this to work here, and agreeing on it with the buyer. That is a thesis. What the analysts’ maps are good for When a company faces a change, it does what the industry has trained it to do. It looks for best practices, and it reads the analysts. The best known of their tools is Gartner’s Magic Quadrant. It plots the vendors in a market on two axes, ability to execute and completeness of vision, which splits them into four groups: Leaders, Challengers, Visionaries and Niche Players. Forrester publishes its own assessment, the Forrester 56

Page 59 of 214. Selling When There Is No Reference Customer Wave, which scores vendors against a set of weighted criteria. IDC has the MarketScape. Alongside them sit peer review sites such as Gartner Peer Insights and G2, where customers rate what they bought, and the bench‐ marks vendors publish against each other. Each of these does a real job. It saves a buyer months of research. It gives a buying committee a shared view from outside. It lets a careful executive point to an independent source. All of them describe a market that has already formed. An analyst can place a vendor only once the category exists, the vendors in it can be compared, and customers have used the products long enough to report back. That takes time. All of it is sensible when someone has gone before you. When the categories move faster than the maps Enterprise AI does not wait for the map. Its categories now form and expire faster than anyone can map them. First came the chat assistant. Then AI that writes and runs code. Then agents that carry out work in many steps. Then whole work surfaces where one person directs many agents. Then personal bots that run all the time and act on your behalf. Each felt like the category until the next one arrived, months later. I name no products here on purpose. By the time you read this, the names would prove the point. So the usual way of deciding has no map to read. Look around, see where the market is heading, follow the leaders in the quadrant: none of it works when the quadrant is drawn after the market has moved. Remember the starlings from the Introduction: thousands of birds turning together with no map and no leader, on a few shared rules. A market with no map moves the same way. Buyers and sellers who hold the same few beliefs can move together without anyone drawing the route. Those beliefs are the thesis. It is the compass the buyer needs, and it is the one thing a seller can bring that no quadrant can. 57

Page 60 of 214. Thesis-Based Selling Where AI sits on the Hype Cycle Gartner has one more chart that almost every technology buyer has seen. In 1995 a Gartner analyst, Jackie Fenn, described a pattern she kept seeing in new technologies and drew it as a curve. She called it the Hype Cycle. It has five phases. An Innovation Trigger gets a technology noticed. Expectations climb to a Peak of Inflated Expectations, where every com‐ pany announces a plan. The results disappoint, and the curve falls into the Trough of Disillusionment. Then, more quietly, people learn what the technology is good for and climb the Slope of Enlightenment, until it set‐ tles on a Plateau of Productivity, where it is simply used. PEAK OF INFLATED EXPECTATIONS PLATEAU OF PRODUCTIVITY SLOPE OF ENLIGHTENMENT enterprise AI buyers INNOVATION TRIGGER TROUGH OF DISILLUSIONMENT Figure 5 The Hype Cycle: expectations over time. Many buyers now stand in the trough. Gartner put generative AI at the Peak of Inflated Expectations in 2023. In its 2025 Hype Cycle for AI, generative AI had slid into the Trough. That is where many of your buyers stand. They have heard that the hype was overdone, and they have seen the pilots that went nowhere. The natural move in the trough is to wait for proof, which means waiting for the plateau, when the references exist. Meanwhile the capability keeps climbing. A buyer who waits for the plateau’s references buys what everyone else bought, a few years late. A thesis is how a buyer climbs out of the trough and acts on the slope, before the plateau arrives. The thesis names what has to be true, and it lets a company move while the evidence is still being made. For you as a seller, that changes the conversation. You stop trying to talk a disappointed buyer back 58

Page 61 of 214. Selling When There Is No Reference Customer up to the peak, which never works, and start helping them agree on the few things that would have to hold for the next step to pay off. The curve has its critics. It describes how people feel about a technology, not what the technology can do, and many technologies never follow it neatly. It is still the clearest picture I know of where enterprise AI buyers stand today. Waiting in the trough for the plateau’s best practices is a reasonable instinct. It is what a company does when someone has gone before it. The question is what to do when nobody has, and the best answer I know comes from one of the boldest projects of the last century. Why best practices fail in a new market Imagine the people who set out for the moon asking for the best practices for going to the moon. There were none. When President Kennedy set the goal in May 1961, the United States had fifteen minutes of experience putting a person in space. Nobody could be called for a reference. The people who went had to write the best practices themselves. So they worked the way a mathematician works. They set down what had to be true about the rockets, the orbits and the human body, secondguessed it, tested each piece on the ground and then in smaller flights, and only then trusted what was built on it. Mercury put one person in orbit. Gemini proved that two spacecraft could meet and dock and that people could live in space for two weeks. Apollo built on both. Each program was a test of the foundations the next one needed. Nobody proves that one plus one is two by finding a customer who tried it. You establish the foundation, and if the foundation is true, what you build from it is true. That is the position your buyer is in. Adopting AI today has more in common with the moon program than with a software upgrade. There is no flight log to copy, and the people in the room will be the first to fly this particular route. What they can do is what NASA did: agree on what has to be true, test it in small steps, and build on what holds. 59

Page 62 of 214. Thesis-Based Selling “ A company adopting AI today is closer to the moon program than to a software upgrade. “Who else is doing it?” is the wrong tool for a trailblazer, because the honest answer is the people in the room. Symmetric uncertainty: neither side has seen how this ends Before I show what the trailblazer does instead, I want to name the position both sides of the table are now in, because the rest of the book is built for it. Put the three shifts side by side: the consumerized C-suite from the Intro‐ duction, the procurement time trap from Chapter 2, and the missing refer‐ ence this chapter has just described. Each one takes away part of what the seller used to bring. Together they change who knows what in the room. Economists have a phrase for a deal where one side knows more than the other: asymmetric information. George Akerlof made it famous in 1970 with a paper about used cars, in which the seller knows whether the car is a lemon and the buyer cannot tell. Enterprise software ran on a friendlier version of the same imbalance. The vendor had watched the category at dozens or hundreds of customers. It knew which rollouts stalled, which configurations held up, and what the third year of ownership felt like. The buyer was doing it for the first time. So the seller knew more, and a good seller earned trust by sharing what they knew. Now look at the moon program again. Nobody at NASA in 1961 knew more about landing on the moon than anybody else, because nobody had done it. Enterprise AI is in that position. The seller still knows more about the product, and has probably seen more pilots than any one buyer has. But on the question that decides the deal, what work inside this company will 60

Page 63 of 214. Selling When There Is No Reference Customer look like in five years and which bets will have paid, the seller is guessing too. Neither side has seen the movie to the end. I call that symmetric uncertainty: not equal ignorance about everything, but equal uncertainty about the one thing that matters most. In my experience most sellers have not adjusted to it. They still walk in with the posture of the side that knows, and the consumerized C-suite from the Introduction can tell, because it has been testing the same AI models at home. A seller in symmetric uncertainty cannot win by claiming to know the ending. The seller’s job is not to pretend to have certainty. It is to bring a point of view you can defend, test it with the buyer, and build conviction together: say plainly what you believe, why, and what would prove you wrong, and invite the buyer to examine it and sign it with you. When there is no precedent to stand on, people buy the thesis before they buy the product. That move has a name in this book, and the next section gives it. A first principle becomes a thesis when the room agrees to it The standard advice at this point is to go back to first principles. I agree, and I think it is half an answer. Here is the other half, said plainly, before I show why it matters: a thesis is a first principle that you put on the table, where everyone in the deal can examine it and agree to it. First-principles thinking is private. A thesis is shared. When you reason from first principles you still look around for one example to check yourself against. In AI that example is the very thing that is missing. You are doing the math without an answer key. You are also not reasoning alone. Gartner’s research on business buying puts six to ten people in a typical buying group, and every one of them arrives with first principles of their own. Picture two of them. The chief financial officer reasons from one premise: the cost of running the business has to fall. From there the logic is clean. Pick the platform that is cheapest to run at scale, consolidate on it, and retire the rest. The chief technology officer reasons from another premise: we will never again be locked into one vendor. From there the logic is just as clean. Pick the open option, keep every part replaceable, and pay more if that is the price of freedom. 61

Page 64 of 214. Thesis-Based Selling Both of them are careful, and every step each of them takes is sound. That is why each one feels rigorous. Given where they started, they are right. And they arrive at opposite vendors. They can argue for a month and neither will move, because they are arguing about conclusions. The disagreement does not live there. It lives in the premises, and nobody has said the premises out loud. Every round of the argument tests the logic, which holds on both sides, and never touches the starting points, which is where they differ. A first principle that lives in your head settles nothing in a room. What the room needs is a short list of statements that everyone at the table has heard, examined and accepted as true: common truths, said out loud, written where all can see them. Once the chief financial officer and the chief technology officer have both agreed that cost has to fall and that no single vendor should hold the keys, they can look for one answer that honors both. Only then can anybody build. IN TWO HEADS YOU BUYER principle A principle B conclusion A conclusion B ON THE TABLE YOU BUYER THESIS said, examined, accepted one conclusion NEITHER MOVES BOTH CAN BUILD Figure 6 Private principles give two conclusions. A thesis gives the room one. Getting the room to that list is the enterprise seller’s job. Finding your own first principles is the easy half. The seller is often the one person who 62

Page 65 of 214. Selling When There Is No Reference Customer talks to every stakeholder, from the chief financial officer to the engineer who will run the system, and the one person whose job is to get the deal to a decision. Nobody inside the buyer’s company owns the agreement. The seller can. Getting that agreement is harder than it sounds. Premises are rarely spoken, and people hold them without noticing. Some are political, tied to a budget or a reputation. People defend positions they have already taken in front of their peers. And the seller has an incentive of their own, which the room can sense: every seller would like the agreed truths to point at their product. A seller who steers loses the room. That is why the rest of this book exists. Parts Two and Three give you the method for finding the theses, getting the room to agree on them, and building the sale on top. FIELD NOTE 04 A first principle is private until someone else agrees with it. A thesis is a first principle you put on the table. 63

Page 66 of 214. Thesis-Based Selling YO U R TAS KS What to do this week Date your stories. Take the three customer stories you use most. Next to each, write when the customer made the decision, and what AI model and price it was made on. Ask whether that customer would make the same choice today. Sort them. Mark each story as production or pilot. Count how many are in production with a measured result. Answer the question straight, once. The next time a buyer asks who else is doing this, say what is true, including what you do not have. Watch what happens in the room. Write one sentence. Finish this line: One thing I believe about this market that would still be true if my company disappeared tomorrow is… Keep it. Chapter 5 will test it. 64

Page 67 of 214. Chapter Four What Every Great Sales Method Quietly Assumes “Those who cannot remember the past are condemned to repeat it.” George Santayana IF THIS IS TRUE Every selling framework solved the bottleneck of its own decade. THEN A new bottleneck calls for a new layer, and the older frameworks keep their jobs on top of it. 65

Page 68 of 214. Thesis-Based Selling So far the argument has been about the ground: the old path, the thesis, and the reference that has gone missing. A seller reading this has a fair question. I have been trained in MEDDIC, or SPIN, or Challenger, and it works. Do I throw it away? No. This chapter walks through the great sales methods one at a time, shows what each was built to supply, and finds the one thing all of them quietly assume. That assumption is exactly where Thesis-Based Selling fits. Then it sets this book beside the three ideas an experienced revenue leader is most likely to hold it up against, The Challenger Sale, Andy Raskin’s strategic narrative and April Dunford’s Sales Pitch, and says plainly where each one ends and this one begins. Enterprise sellers are a particular breed. The job is to change people’s minds, on decisions that cost millions, in rooms full of people who would rather not change anything. It takes interpersonal skill to read the room and technical skill to stand behind every claim. The quota resets every quarter. Few people last long in the job, and the ones who do tend to be driven, rigorous and restless to get better. That is why so many sellers read. Look at the shelf of anyone who has sold for a decade and you will find the books: on negotiation, on persuasion, on psychology, on how to run a deal. Sellers work on themselves because the job punishes anyone who stands still. It is also why sellers, and the people who lead them, lean on frameworks. A framework gives a team one language, so that when a manager asks whether a deal has a champion, an insider who will sell for you when you are not in the room, everyone means the same thing. It makes good habits repeatable, so a new hire can run a deal the way the best people do. It gives a manager something to coach against and a forecast to inspect. Whole profes‐ sions, sales training and sales enablement, exist to teach the frameworks and run them. Chapter 1 walked the proven path those frameworks were built for. This chapter looks at the frameworks themselves: BANT, Consultative Selling, Solution Selling, Strategic Selling, SPIN Selling, MEDDIC, The Challenger Sale, Gap Selling and The JOLT Effect. Each one was an answer to some‐ thing that had become scarce in its time, and each one worked. Walk through them in order and you can watch the scarce thing change. Then look at what all of them assume, because that assumption is the one AI has broken. 66

Page 69 of 214. What Every Great Sales Method Quietly Assumes Sales frameworks over the years, and what each one supplied Here is the lens for this section. Each framework was built for whatever was scarce in deals in its era, and it worked by supplying that thing. Read them in order and you are reading the history of the profession: every few years, something new ran short, and someone wrote down how the best sellers were getting it anyway. BANT: the seller’s time. The oldest is not a method at all. It is a filter. BANT is attributed to IBM, usually dated to the 1950s or 1960s, though no document from that time has surfaced and no author is known. The letters ask four questions of a prospect: is there a Budget, is there someone with the Authority to buy, is there a Need, and is there a Timeframe? (IBM’s own partner forms from the 2000s say Timeframe; most people now say Time‐ line.) In the mainframe era a deal was expensive and slow, and a field seller could chase only a few at once. What was scarce was the seller’s time. BANT spent it only where money, a decider, a need and a date already existed. It says nothing about how to create a need, and it tends to reject exactly the deals a new category produces: no budget line yet, no clear owner, a need nobody has named, no date. Consultative Selling: a seat at the top. In 1970 the American Manage‐ ment Association published Consultative Selling, by Mack Hanan with James Cribbin and Herman Heiser. Hanan saw sellers stuck “vending”: selling products and prices to purchasing agents, watching margins fall. What was scarce was credibility with the people who ran the customer’s business. His answer was to sell an improvement in the customer’s profit, not a product. The seller researched the customer’s operations, called on the managers who owned the numbers, and wrote a Profit Improvement Pro‐ posal that said how much profit, how soon, and how sure. The idea spread so widely that “consultative” came to mean little more than asking questions and being helpful, which lost Hanan’s hard arithmetic. And the “how sure” leaned on something the book took for granted: results from earlier cus‐ tomers to estimate from. Solution Selling: a picture of the fix. Michael Bosworth sold for Xerox Computer Services from 1972, designed its sales training, and worked with Neil Rackham on the research Xerox sponsored into large sales. In 1983 he founded Solution Selling, and in 1994 he put it in a book. He was selling 67

Page 70 of 214. Thesis-Based Selling systems that most buyers did not know they needed. What was scarce was the buyer’s clarity about the need. Bosworth described levels of need: latent pain the buyer does not see or will not admit, pain the buyer admits, and a vision, where the buyer can picture the solution. The seller’s job was to move the buyer up those levels and shape the vision so that it matched what the seller could do. The method came with tools: a nine-block grid of questions for building the vision in discovery, the pain chain that traces one person’s problem to another’s, and a sponsor who opens the door to the person with power. It assumes a problem the seller can name and a solution category the buyer can picture once it is described. Strategic Selling: a map of the buyers. By the 1980s large purchases had several people in them, and sellers sold to whoever would take the meeting. Robert Miller and Stephen Heiman, who founded their firm in 1978, published Strategic Selling in 1985. What was scarce was visibility into the buying organization. Their method maps the buying influences by role, not title: the Economic Buyer who gives final approval and controls the money, the User Buyers who will live with the product, the Technical Buyers who screen it against specifications and can say no but not yes, and the Coach who wants you to win and guides you. For each influence the seller writes down their mode (looking to grow, in trouble, on an even keel, or overconfident), the business result and the personal win they want, and the red flags. All of it goes on one page, the Blue Sheet, which became the agenda of the deal review. Strategic Selling assumes a buyer that has already decided to evaluate something, with roles that can be named. SPIN Selling: urgency in large sales. Neil Rackham’s SPIN Selling (1988) is the one framework of the era built on large-scale research. His firm, Huthwaite, studied 35,000 sales calls over twelve years in more than twenty countries, with sponsors that included Xerox and IBM. The training of the day, closing techniques and objection handling, had come from small sales. Rackham found it failed in large ones. What was scarce in a large sale was the buyer’s urgency: the buyer had problems, but did not value solving them enough to take a large, risky decision. SPIN supplies it through questions in a rough order. Situation questions gather the facts, and successful sellers ask fewer of them. Problem questions uncover difficulties. Implication ques‐ tions explore what those difficulties cost, and they were the strongest predictor of success in large sales. Need-payoff questions let the buyer say, in their own words, what solving the problem would be worth. Huthwaite 68

Page 71 of 214. What Every Great Sales Method Quietly Assumes reports that the first thousand people trained sold 17 percent more than control groups, and Rackham has been clear that the gain came from the whole coaching process, “not the four little questions.” SPIN assumes the seller’s offer is a known way to solve the problem, and that the seller can show it works. MEDDIC: the truth about the forecast. In 1996, at the software company PTC, Dick Dunkel and Jack Napoli, working for the sales leader John McMahon, asked three questions of their deals: why do we win, why do we lose, and why do deals slip? The same six factors kept appearing, and Dunkel wrote them on a whiteboard: Metrics, the measurable gain the buyer expects; the Economic buyer, the one person who can say yes when everyone else says no; the Decision criteria and the Decision process; the pain the deal solves, Identify pain; and a Champion, an insider with power who sells for you when you are not in the room. Enterprise software deals were large and slow, and forecasts were built on sellers’ optimism. What was scarce was a shared test of whether a deal was real. MEDDIC turned opinion about a deal into evidence a manager could inspect. PTC passed a billion dollars in revenue by the end of the 1990s, and in the account its sales leaders give, MEDDIC was a large part of why, and the company then went years without missing a forecast. Practitioners later added a second C for Competition and a P for the Paper process, the path from a verbal yes to a signed contract, making MEDDPICC. MEDDIC qualifies a deal; it does not tell you how to create the pain or the criteria. And its Metrics are easiest to fill when someone has already measured the gain elsewhere. The Challenger Sale: insight. During the downturn of 2008 and 2009, researchers at CEB, the Corporate Executive Board, had managers rate more than 6,000 sellers on 44 attributes. Matthew Dixon and Brent Adamson published the result in 2011 as The Challenger Sale. The sellers clustered into five profiles: the Hard Worker, the Relationship Builder, the Lone Wolf, the Reactive Problem Solver and the Challenger, who has a different view of the customer’s world and likes to debate it. About 40 percent of star performers were Challengers, and in complex sales the share was higher. Relationship Builders, the profile many managers hired for, were 7 percent of the stars. Buyers were now doing much of their research before they called a vendor. What was scarce was insight the buyer did not already have. So the Challenger teaches the customer something new about their own business, tailors it to each person in the room, and takes control of the conversation, 69

Page 72 of 214. Thesis-Based Selling including the conversation about money. The study was rated by managers and showed correlation, not cause, and academics have criticized it for that. But the idea stuck. Its teaching pitch leans heavily on “what we see at compa‐ nies like yours,” which is precedent by another name. Gap Selling: a diagnosis. By the late 2010s product information was everywhere, and sellers led with demos. Deals stalled because nobody had diagnosed the real problem. In 2018 Keenan, who goes by one name, published Gap Selling. What was scarce was diagnosis. His framework has three parts: the current state, which covers the buyer’s environment, the problem, its impact, its root cause and how people feel about it; the future state the buyer wants; and the gap between them. The bigger and clearer the gap, the more value and urgency. Gap Selling assumes the buyer can describe where they are and believe in where they want to be, and that future state is usually made believable by pointing at someone who got there. The JOLT Effect: the courage to decide. In 2022 Matthew Dixon and Ted McKenna analyzed more than 2.5 million recorded sales conversations for The JOLT Effect. Between 40 and 60 percent of qualified deals ended in no decision at all. The buyer agreed that change was needed and still could not commit, out of fear of making a mistake they would be blamed for. What was scarce was the buyer’s confidence to decide. The authors traced indecision to three sources: trouble choosing between options, a need for more information, and fear that the promised result will not arrive. JOLT names four moves: Judge the indecision, Offer your recommendation, Limit the exploration, and Take risk off the table. A recommendation like “cus‐ tomers like you usually choose this” and the risk reducers that back it depend on knowing how similar customers fared. That is seventy years of the profession in nine methods. Each one answered a real scarcity, and each one worked. The table puts them side by side. 70

Page 73 of 214. What Every Great Sales Method Quietly Assumes When Framework What deals lacked, which it supplied What it assumes is already there 1950s? BANT The seller’s time: which deals deserve effort A budget line and a known need 1970 Consultative Selling A seat with the people who run the business Earlier results to estimate the profit from 1983 Solution Selling The buyer’s clarity about the need A solution the buyer can picture 1985 Strategic Selling A map of who decides and what each wants A buying process with roles to name 1988 SPIN Selling Urgency in a large sale A fix already known to work 1996 MEDDIC A shared test of whether a deal Metrics measured at earlier is real customers 2011 The Challenger Sale Insight the buyer did not have Patterns seen across many customers 2018 Gap Selling A diagnosis of the real problem A future state someone has reached 2022 The JOLT Effect The buyer’s confidence to decide Knowing how similar customers fared Now Thesis-Based Selling A reference: proof it works for someone like you Truths agreed in the room Figure 7 Sales methods by era: what each supplied, and what each assumes. How to read the table: each row is one method and the year it appeared. The third column is what was in short supply in deals at the time, which the method supplied. The fourth is what the method takes for granted. That last column is where AI does its damage. The acronyms: BANT is Budget, Authority, Need, Timeframe. SPIN is Situation, Problem, Implication, Need-payoff. MEDDIC is Metrics, Economic buyer, Decision criteria, Deci‐ sion process, Identify pain, Champion; MEDDPICC adds Paper process and Competition. JOLT is Judge the indecision, Offer your recommenda‐ tion, Limit the exploration, Take risk off the table. BANT’s date is uncer‐ tain, hence the question mark. 71

Page 74 of 214. Thesis-Based Selling The last row is the odd one out, and it is the point of the table. What deals lack today is the very thing every row above it took for granted: a refer‐ ence, proof that the product works for someone like this buyer. Thesis-Based Selling does not supply that proof, because nobody can yet. It supplies a substitute the room can stand on, a few truths agreed by the people in the deal. When you sit in your next deal review, read your own method’s row and ask whether its last column is really there in this account. The next section makes that test explicit. The one thing every method needs: precedent Read the last column of that table. It is the same thing nine times, in different words: precedent. BANT assumes a budget line already exists, which means the category has been bought before. Consultative Selling’s “how sure” is an estimate from earlier customers. Strategic Selling assumes a buying process whose roles can be named, which is easy when the company has bought this kind of thing before and hard when nobody owns it yet. SPIN and Solution Selling assume the fix is a known quantity. The seller can afford to ask good ques‐ tions because the answer is already on the shelf. MEDDIC starts with Metrics, the measured benefit, and that measurement comes from earlier customers. The Challenger’s teaching insight comes from seeing the same pattern across many accounts, which is another way of saying references. Gap Selling needs a future state the buyer finds believable, and it is believable because someone got there. JOLT’s last move, taking risk off the table, is usually done with proof. None of the authors hid this. They did not need to say it, because in their markets precedent was always there. In enterprise AI it is the missing input. That is the opening Thesis-Based Selling fills. 72

Page 75 of 214. What Every Great Sales Method Quietly Assumes Each framework is a way to run a sale across ground that somebody has already walked. In AI, hardly anybody has walked it. The methods still work as methods. They have less to stand on. When the seller no longer knows more than the buyer Of all the methods in the table, The Challenger Sale is the one this book will be compared with most often, and it deserves a careful answer, because it is close to this book and I owe it a great deal. Dixon and Adamson called the heart of their method commercial teaching. The Challenger does three things: teaches the customer something new about their own business, tailors the message to each person in the room, and takes control of the conversation. The teaching follows a choreog‐ raphy the book lays out in six steps. A warmer shows that you understand the customer’s world. A reframe tells them something they had not consid‐ ered. Then comes what the authors call rational drowning, the numbers that show the reframe is costly, then the emotional impact on the people in the room, then a new way of acting, and only at the very end your solution. The authors are firm that the insight has to lead back to something the supplier does better than anyone else. That is good selling, and on a mature product I would want every seller I lead to do it well. Now look at where the insight comes from. In their examples it comes from what the supplier has seen across its customers: the hidden cost most companies miss, the pattern that only shows up at the hun‐ dredth deployment. Commercial teaching is built for asymmetric infor‐ mation. The vendor has watched the movie many times, the buyer is watching it for the first time, and the seller’s job is to tell the buyer how it tends to end while there is still time to change the plot. The seller’s authority in the room comes from that wider view. 73

Page 76 of 214. Thesis-Based Selling Symmetric uncertainty, from Chapter 3, takes the wider view away. Commercial teaching asks the seller to bring an insight the buyer had not considered. What happens when neither side has enough evidence to know what is right? When neither side has seen the movie five years out, a seller who teaches as if they had is bluffing, and the consumerized C-suite will find the bluff. So Thesis-Based Selling keeps the Challenger’s courage, the will‐ ingness to tell a buyer something they did not expect to hear, and changes what the seller brings. Instead of a lesson the vendor already knows, the seller brings a belief about where the technology is going, says what would prove it wrong, and asks the buyer to examine it and sign it with them. I call that shared thesis formation, to set it beside commercial teaching, and four differ‐ ences separate them. Who knows more. Commercial teaching assumes the seller does. Shared thesis formation assumes both sides are uncertain about the future, and says so out loud. Which way the idea flows. Teaching flows from the seller to the buyer. A thesis is formed between them: the seller proposes it, the buyer corrects it, and the corrected version belongs to both, as Chapter 9 shows with Priya. What makes it trustworthy. A teaching insight earns trust from the vendor’s experience. A thesis earns it from its falsifier, the agreed statement of what would prove it wrong, which both sides can check on a date they set together. Where it leads. A Challenger insight is designed to lead to the supplier’s unique strengths. A thesis has to be true whichever vendor wins, or it is a pitch in disguise; the product appears only at the top of the stack. Hear the difference in one opening each. The Challenger seller says: “Companies like yours lose more analyst time to rework than they ever measure, and here is what the best of them do about it.” That is a strong line when the seller really has seen hundreds of companies like yours. The thesis seller says: “We believe your definitions will outlive any AI model you buy. Nobody has run this long enough to prove it, so here is what would show we are wrong, and here is how we could check it together in the first quarter. Where do you disagree?” The first asks the buyer to trust what the vendor has seen. The second asks the buyer to examine a belief and own it. I want to be fair to Challenger here, because the two are not enemies. Where a market has references, commercial teaching is the right tool, and 74

Page 77 of 214. What Every Great Sales Method Quietly Assumes many real deals use both: a reframe opens the meeting, and the thesis is what the reframe becomes once the buyer has pushed on it and agreed. When I sell the mature parts of a platform, I teach. When I sell the part nobody has run for five years, I form the thesis with the buyer. Your story is not the buyer’s bet: strategic narrative and the sales pitch Two more ideas sit on the same shelf, and a sales leader who knows them will ask how this book differs. Both are about the vendor’s own story, and both are very good at it, so it is worth placing them honestly. The first is Andy Raskin’s. In 2016 Raskin, a messaging consultant who had run marketing at venture-backed companies, published an essay on Medium called “The Greatest Sales Deck I’ve Ever Seen,” about a sales deck from Zuora, the subscription billing company. It was shared widely, and Raskin has since built a practice on the idea, which he calls the strategic narrative. Do not open with the product. Open with a big, undeniable change in the world, for Zuora the shift to what it called the subscription economy. Show that the change will create winners and losers. Tease the promised land the winners reach. Present your features as the magic gifts that help the buyer get there, and finish with evidence that you can make the story come true. The second is April Dunford’s. Dunford, known for her book on posi‐ tioning, Obviously Awesome (2019), published Sales Pitch in 2023. She argues that a business-to-business pitch should open with an insight about how the market has changed, then walk fairly through the alternatives the buyer could choose, with their strengths and their limits, then describe what a perfect solution would look like, and only then show the product’s differ‐ entiated value and the proof behind it. Her aim is a buyer who can decide with confidence, not a buyer who has been dazzled. Both start from a change in the world instead of the product, and so does this book, so a seller trained in either will find the bottom layer of the stack familiar. Three things separate them from a thesis. The first is who owns it. The narrative and the pitch belong to the vendor: written by the vendor, polished by marketing, and told to every buyer in much the same 75

Page 78 of 214. Thesis-Based Selling ned it and signed it. The second is whether it can be proved wrong. A narra‐ tive is built to persuade, so it does not say what would sink it, and nobody checks the promised land at the first quarterly review. A thesis names its falsi‐ fier and invites the check. The third is what happens next. A story is told and retold. A thesis is defended, revised in the open, and written into the buyer’s own decision, as Chapter 9 shows. So use them together. Raskin’s big change and Dunford’s market insight are often where a good thesis starts. The step this book adds comes after the story: hand the change to the buyer as a claim, ask where they disagree, write down what would prove it wrong, and build only on what both sides have signed. Your story persuades the buyer. The thesis is something the buyer owns. A thesis and a hypothesis are different things One last idea sits closest of all, and it comes from the Challenger tradition itself, so I want to separate it carefully, because the words sound alike and the jobs are different. Sellers trained on Challenger are often taught hypothesisled selling: research the account, walk in with a guess about the customer’s problem, and test the guess in the meeting. That is a good habit. Here is what each word means in this book. A hypothesis is the seller’s guess about one account’s problem. I think your analysts spend too much time cleaning data. You bring it to a discovery call to find out whether you are right. A thesis is a belief about how the world works, stated plainly and agreed on by the people in the deal. Our data will outlive any tool we run on it. You set it down so that a decision can be built on it. The differences follow from that. Who owns it. A hypothesis belongs to the seller. The buyer may never hear it said out loud. A thesis belongs to the room. It is only a thesis in the sense of this book once the buyer has examined it and agreed. What it is for. A hypothesis is a tool for finding the right problem. You test it, and if it is wrong you throw it away and try another. A thesis is a foundation. You build on it: an implication, a decision, and last of all a product. 76

Page 79 of 214. What Every Great Sales Method Quietly Assumes How long it lives. A hypothesis lives for a meeting or two. A thesis lives for the life of the deal and beyond it. You carry the same few theses from account to account, and they get sharper each time someone attacks them. What happens when it is wrong. A hypothesis that fails is simply replaced, and nothing is lost. A thesis that turns out wrong is revised in the open, with the buyer, and everything built on it is checked again. That is why a thesis has to name what would prove it wrong. Here is one account seen both ways. A seller walks into a manufacturer with a hypothesis: your reporting team is drowning in manual data work. Discovery shows it is half right. The real pain is that every new tool the company buys needs its data copied and reshaped again. So the seller drops the first guess and finds the problem. Then, with the chief information officer and her architects, the seller writes down something they all believe: our data will outlive any tool we run on it, so it should sit in an open format any engine can read. Nobody discards that sentence after the meeting. The architecture decision is built on it, the vendor shortlist follows from it, and it will still be true at the next company the seller sells to. The hypothesis found the problem. The thesis decided the deal. “ You can lose a hypothesis in every meeting and still hold the same three theses all year. How Thesis-Based Selling works alongside the methods you already use The Thesis-Based Selling framework replaces none of these methods. It sits underneath them. Here is each one, for a reader who has never been trained in it, with where it sits in a deal and where the thesis plugs in. MEDDIC, and its longer form MEDDPICC, for qualification. MEDDIC is a checklist for deciding whether a deal is real: Metrics, the measurable benefit; the Economic buyer, the person who can spend the money; the Decision criteria and the Decision process the buyer will use; the pain the deal solves; and a Champion inside the buyer who will 77

Page 80 of 214. Thesis-Based Selling fght for it. MEDDPICC adds the Paper process, how the contract gets signed, and the Competition. It runs through the whole deal and is reviewed in every forecast call. The thesis plugs into the D’s and the M. Decision criteria only mean something once the buyer has agreed what is true: criteria written before the thesis tend to be copied from the last purchase. In AI the Metrics rarely come from earlier customers, so they come from the thesis instead: if this is true, here is what we should be able to measure. The Economic buyer is the person who must hold the foundation thesis, in their own words. Identify pain becomes the impli‐ cation: what the thesis means for this company, the second of the four layers Chapter 9 builds. The Champion’s real job becomes carrying the thesis into rooms you are not in. And Competition now includes the thesis the alternative rests on, doing nothing among them. SPIN Selling, for questions. SPIN orders the questions in discovery: Situation questions about the facts, Problem questions about difficulties, Implication questions about what those difficulties cost, and Need-payoff questions that let the buyer say what solving them would be worth. It lives in the early meetings. The thesis plugs into the last two. Implication questions are how you find the beliefs under a buyer’s position: ask what follows from a problem, and the buyer tells you what they believe about the future. Needpayoff questions are how the buyer comes to say the “then” of a thesis in their own words, which is worth more than hearing you say it. Strategic Selling, for the people. Miller and Heiman taught sellers to map every buying influence: the economic buyer, the users, the technical gatekeepers and the coach who guides you. It is the map you keep for the whole deal. The thesis plugs in here because each of those people arrives with premises of their own, like the chief financial officer and the chief tech‐ nology officer in Chapter 3. The map tells you who has to agree. The thesis is what they have to agree on. Solution Selling, for the pain and the vision. Bosworth’s method finds the buyer’s pain and then builds, with the buyer, a vision of what solving it looks like. It sits in discovery and in the proposal. In AI the pain is often unformed, because the buyer has not yet lived with the problem long enough to name it. The thesis gives the vision something to stand on: here is what we believe about where your business is going, so here is what solving it looks like. 78

Page 81 of 214. What Every Great Sales Method Quietly Assumes The Challenger Sale, for insight. The Challenger teaches the customer something new about their own business, tailors it to each person, and takes control of the conversation. It sits in the first serious meeting. Where a market has references, teach. Where it does not, form the thesis with the buyer, as this chapter described earlier: the reframe becomes the moment you lay the thesis on the table, and taking control means keeping the room on the thesis instead of drifting back to features. Gap Selling, for the size of the change. Keenan’s method measures the distance between the customer’s current state and the future state they want, and sells the value of closing it. It sits after discovery, when you build the business case. The future state is a thesis about the future. Agree on it first, and the gap you measure is one the buyer believes in. Skip that step, and the buyer argues with your future state instead of your price. The JOLT Effect, for indecision. Dixon and McKenna’s four moves get a frozen buyer to a decision: judge the indecision, offer a recommendation, limit the exploration, take risk off the table. They belong at the end of a deal that has stalled. In AI much indecision comes from a thesis nobody agreed on: the buyer is afraid of being wrong about the future. So the recommenda‐ tion should come from the agreed thesis, and the thesis also limits the explo‐ ration, because it narrows the field to the products that fit it. The risk can be taken off the table without a reference: name what would prove the thesis wrong, and set an early checkpoint where you will both look. WITH REFERENCES WITHOUT THEM THE FRAMEWORKS MEDDIC, SPIN, Challenger, JOLT THE FRAMEWORKS MEDDIC, SPIN, Challenger, JOLT EARLIER RESULTS what other customers got THESES agreed in the room, deal by deal Figure 8 The frameworks keep their jobs. What they stand on changes. The figure shows the idea in one picture. At the top, the methods keep their jobs: MEDDIC still qualifies, SPIN still asks, Challenger still teaches, JOLT still closes. Underneath, the ground changes. In a market with refer‐ ences, the methods stood on earlier results. In enterprise AI they stand on a 79

Page 82 of 214. Thesis-Based Selling few truths agreed in the room, deal by deal. Nothing about the methods has to be unlearned. What has to be learned is how to build the ground they now stand on, which is what Parts Two and Three teach. Here is one deal with all of them in it. You will meet the thesis session, the stack and the ledger properly in Part Three; for now, notice where each method sits. 1 Qualify. In the first week, run MEDDPICC. Is there a decision to make, an economic buyer and a reason to act this year? 2 Agree inside. Before the first big meeting, run the thesis session from Chapter 8 with the person who sells beside you. 3 Discover. Use SPIN. Let the Implication questions surface what the buyer believes. Map each buying influence and their premises, Strategic Selling style. 4 Reframe. In the first serious meeting, lead with a Challenger insight: lay down three theses and stipulate what everyone accepts, as Chapter 9 describes. 5 Build. With the buyer, build the stack, the chain from thesis to implication to decision that Chapter 9 describes. Write the Decision criteria from it. Size the gap, Gap Selling style, against the agreed future state, and shape the solution vision on it. 6 Carry. Give your Champion the stack in three sentences, so it travels without you. 7 Defend. Take every challenge as Chapter 10 describes, and keep the ledger. 8 Decide. If the deal stalls, use JOLT: recommend from the thesis, narrow the options, and take the risk off the table with the falsifier, the agreed statement of what would prove the thesis wrong, and an early checkpoint. 80

Page 83 of 214. What Every Great Sales Method Quietly Assumes Look at what the list shows. Every method you already know still has a step, and in roughly the place you would expect: qualification first, discovery early, insight in the first serious meeting, the business case in the middle, the push to decide at the end. What is new is steps two, five and six, and all three are about the thesis: agreeing on it inside your own team, building the deal on it with the buyer, and making sure it travels without you. If you run only those three on your next deal and leave everything else as it is, you will already be selling differently. In a sales organization the work divides naturally. Sales engineering leaders own the thesis session. Enablement owns the team’s ledger and the practice of testing theses against an AI model before they reach a buyer. Operations adds a few fields to the CRM, and the forecast call keeps MEDDPICC and gains one question about the thesis; the Field Guide at the back of the book gives both, with the scorecard for that conversation. The methods tell you how to run each part of the sale. What changes is the ground. Those methods stood on a stack of earlier results. Now they stand on a small number of statements the seller and the buyer have agreed are true. That agreement has to be built, deal by deal, and building it is a skill the older books did not need to teach. FIELD NOTE 05 The frameworks tell you how to run the sale. The thesis is what you stand on while you run it. 81

Page 84 of 214. Thesis-Based Selling YO U R TAS KS What to do this week Name your framework. Write down which method your team actually runs today, officially or by habit. Be honest: many teams say MEDDIC and run “whatever the forecast call asks for,” and it helps to know which one you are working inside. Find what it leans on. For that method, write the one thing it needs from the outside world to work: prior results, a known fix, a measured benefit at earlier customers. That is the assumption this chapter says AI has broken. Check the supply. For the AI product you sell or buy, ask how much of that thing exists today, and be exact. Count the customers in production with a measured result, not the logos on the slide. The gap between the two numbers is the gap a thesis has to fill. Keep the method. Decide which parts of your current framework still earn their place. Most of them will. You are not replacing your method; you are giving it ground to stand on. 82

Page 85 of 214. Part II Establish the Thesis Round one in practice: what makes a thesis, how to find three to five under a decision, and the five I hold about AI.

Page 86 of 214.

Page 87 of 214. Chapter Five What a Thesis Is, and How to Test One “Here I stand. I can do no other.” Attributed to Martin Luther IF THIS IS TRUE A thesis is a claim about the world that you write down and agree to defend. THEN Anyone can test it, and whatever survives the testing can carry weight. 85

Page 88 of 214. Thesis-Based Selling Part One laid the ground: how enterprise selling was built, why people buy the thesis, why the reference is missing, and where the methods you already use need help. If the book stopped there, you would know what has changed but not what to do about it. Part Two is round one of the framework, establishing the thesis, and it takes this chapter and the next two. Chapter 2 gave the word its plain meaning. This chapter makes it exact: where the word comes from, the tests that tell a real thesis from a slogan or a pitch, where theses come from in a deal, and how many a deal should rest on. It starts with three traditions that have used the word seriously, because a seller can borrow from all three, and the third of them shows why buying AI now looks more like leading a Series A than buying software. Luther’s Ninety-five Theses: a claim posted for debate In the early 1500s the Catholic Church was raising money by selling indul‐ gences: certificates said to reduce the punishment for sins, for the buyer or for a relative already dead. Part of the money went to rebuilding St. Peter’s Basilica in Rome. In 1517 a Dominican friar named Johann Tetzel was selling them near Wittenberg, in Saxony, with a sales pitch that has outlived him: as soon as the coin in the coffer rings, the soul from purgatory springs. Martin Luther, a monk who taught theology at the university in Witten‐ berg, thought the practice was wrong, and that it was teaching people the wrong thing about forgiveness. Plenty of people had complained before him. Complaints had changed nothing. On October 31, 1517, Luther did something more specific. He wrote ninety-five numbered propositions, in Latin, the language of scholarly debate, and sent them to his archbishop. According to tradi‐ tion he also nailed them to the door of the Castle Church in Witten‐ berg, which served as the university’s notice board. Historians still argue about the door. They do not argue about the title, which says exactly what the document was for: a Disputation on the Power and Efficacy of Indulgences. A disputation was a formal academic debate. Luther was posting the claims he was prepared to defend, and inviting anyone to come and argue. 86

Page 89 of 214. What a Thesis Is, and How to Test One What happened next is why the word still carries weight. Printers in several cities copied the theses, translated them into German, and within weeks they were being read across the German lands. The argument no longer depended on who had the most authority. It had moved onto ninetyfive stated claims that anyone could read, accept or contest, one at a time. Four years later, at the Diet of Worms, an imperial assembly called to judge him, Luther refused to take his writings back. The Reformation that followed split Western Christianity. Look at the form, because it is the form this book borrows. The claims are numbered. They are public. Each one can be attacked on its own. The author put his name to them. Disagreement that had gone nowhere for years started to move once it was set down as plain propositions. You have probably sat in a version of that stalled argument. A buying committee has heard every vendor’s opinion about AI, and every vendor has more authority on its own slides than the last. Without a reference customer, nobody can settle it by pointing at someone who already succeeded. What moves the room is the same thing that moved Wittenberg: a short list of claims, written down, that people can take one at a time and either accept or attack. The doctoral thesis: a claim defended before a committee The second tradition is in every university. A doctoral candidate spends years on a thesis: a claim about the world, with the evidence and the argument for it. Then comes the defense, in Britain and elsewhere called the viva, short for the Latin viva voce, “with the living voice.” The candidate stands in front of a committee of experts whose job is to find the hole. They question the evidence, the method and the conclusion, out loud, for an hour or more. The degree is awarded for surviving the questions. Writing the document is only the entry fee. A committee does not reward a candidate for never being challenged; it rewards the candidate for answering the challenge, conceding where they must, and showing that what is left still stands. That is exactly what the Defend step asks of a seller in Chapter 10. You do not protect a thesis by keeping it away from objections. You find out 87

Page 90 of 214. Thesis-Based Selling what it is worth by letting the sharpest people in the room attack it. A thesis that has never been attacked has an unknown strength. The defense is where its value gets measured. The investment thesis: a claim a fund puts money behind The third tradition is the one executives know best, because it decides where venture money goes. Venture investors write an investment thesis, a short statement of what they believe about the future and therefore what they will fund. Some became famous. In 2011 Marc Andreessen set his firm’s belief down in The Wall Street Journal in five words, software is eating the world, and Andreessen Horowitz invested on that belief for years. Here is how a thesis works inside a fund, day to day. The same year, Brad Burnham of Union Square Ventures put his firm’s thesis in a single tweet: invest in “large networks of engaged users, differenti‐ ated through user experience, and defensible through network effects.” One sentence did a lot of work. Partners could say no to a pitch in minutes and yes in days. Founders knew before the first meeting whether they fit. When the world changed, the firm changed the sentence and published the new one, in 2015 and again in 2018. Everyone could see what the partners had believed, what they believed now, and why it moved. An investor with a thesis is doing what a seller in AI now has to do, which is to commit money and reputation to a future that has no reference customer yet. The fund cannot wait for proof, because by the time proof exists the opportunity is priced in. Your buyer is in the same position with AI, and the thesis is what lets both of you commit before anyone else has gone first. Put the three traditions together and you have the meaning this book uses. Luther shows the claim set down in public, where others can examine it. The committee shows it attacked and defended. The fund shows it agreed 88

Page 91 of 214. What a Thesis Is, and How to Test One on and built on, with real money behind it. A thesis is a claim set down to be examined, agreed and built on. Buying foundational AI is like leading a Series A Before the tests, one of the three traditions deserves a second look. The Introduction promised to come back to it, and it describes your buyer’s posi‐ tion more exactly than any comparison with buying software. Start with how venture money works. A young company raises money in rounds. The seed round is small, early money. The Series A is usually the first large round from venture firms, and one firm leads it: it sets the terms, puts in the largest share, often takes a seat on the board, and has to carry the deci‐ sion through its own partners. At that stage the company has a product and some early signs, but no proof of the business it will become. So the lead partner writes an investment memo. It says what the firm believes about the market, why this team, why now, what has to go right, what could go wrong, and what the company must show before it earns the next round. The money comes in stages, and each new round is a checkpoint against the memo. Now put your buyer in that seat. A company committing to founda‐ tional AI, the AI models and platforms its work will be rebuilt on rather than one more tool for one team, is leading a round. It puts in money, and much more than money: its data, months of its people’s time, the rebuilding of workflows that took years to settle, and the reputations of the executives who sponsor it. What it is backing is not a finished product. It is a change in how the company works, which the company and the vendor will figure out together, and its return depends on things nobody can promise, such as how fast the AI models improve and how readily people take to them. It even comes in stages: a pilot that plays the part of the seed round, a first rollout that plays the Series A, and the scale-up that has to be earned like the next round. That is why the legacy software scorecard fails here. A scorecard gives points for the features of a finished product and adds up the columns, because it assumes the question is which finished product fits best. A venture partner who scored young companies only on today’s features would miss nearly every one worth backing, because what matters is where 89

Page 92 of 214. Thesis-Based Selling the company is going and whether the thesis behind it holds. Foundational AI is the same: as Chapter 2 showed, the features a committee scores this quarter will have changed several times before the contract ends. What the decision needs is the partner’s memo: the three to five beliefs the bet rests on, why now, what has to be true, what would prove the bet wrong, the mile‐ stones that release the next stage, and what the company keeps if it stops. I want to be careful with the comparison, because it has limits. Your buyer does not get shares in your company, and you are not asking for a stake in theirs. The parallel is about the shape of the decision, not the money: a large, staged commitment to a future nobody has seen, made by people who will have to explain it later. Within that limit, it changes how you see your own role. When you sell foundational AI, you are less like a vendor filling in a scorecard and more like a founder in a partner meeting. Your sponsor is the lead partner, and the buying committee is the partnership they must go back to and persuade without you in the room. The most useful thing you can do is help them write the memo, and that memo is a thesis. Three tests for a thesis LUTHER, 1517 THE DOCTORATE THE INVESTOR numbered, public, signed defended out loud commits to a future A THESIS a claim about the world that you write down, sign and agree to defend Figure 9 Three traditions, one working definition. The figure lines the three traditions up. Luther’s theses were numbered, public and signed. The doctoral thesis is defended out loud. The investor’s thesis commits money to a future. Each contributes one part of the defini‐ tion at the bottom, and the tests that follow check a sentence against all three parts: can it be examined, can it be defended, and can someone build on it? When you have a candidate, a sentence you think might be a thesis, run it through three tests before you take it into a room. Think of them as a 90

Page 93 of 214. What a Thesis Is, and How to Test One screen. A sentence that fails any one of them is something else: a slogan, a pitch, a requirement or a guess. Each test also does a job later in the deal, which is why it is worth being strict here. 1 Someone reasonable could disagree with it. “AI will change every industry” fails, because no one in the meeting will argue. “Within two years you will swap the AI model under your application at least once” passes. A person can say no to that, and then you have a conversation. This matters because agreement only means something where disagreement was possible. A room that nods at a truism has agreed to nothing, and nothing can be built on it. And the disagreement, when it comes, is where you learn what this buyer believes, as Chapter 9 shows. 2 It stays true whoever wins the deal, and whatever product is chosen. A thesis describes the world, not your product. Read the sentence as if you worked for your competitor. If it stops being true, it was a pitch. This matters because buyers discount anything that only works if they pick you. A thesis that would survive a competitor’s win is one the buyer can adopt without feeling sold, and it stays in the account even if you lose the deal. 3 It names what would prove it wrong. The philosopher Karl Popper drew the line between science and everything else here: a real claim tells you what evidence would sink it. This matters because the falsifier turns a belief into something both sides can check. It gives the trial its measure, and it makes you sound like someone who has thought the claim through, as Chapter 10 describes. A thesis that passes all three tests still has to hold up in a room. Three more things are asked of it there, and all three are easy to underrate. The thesis has to change something The first is that something has to follow from the thesis. If the buyer agrees with it and nothing in their plan changes, the thesis is trivia. “Data matters” is true, and a room will agree with it, and the next morning every budget is 91

Page 94 of 214. Thesis-Based Selling exactly where it was. Compare “our data will outlive any tool we run on it.” If the buyer agrees, something changes at once: the next tool they buy has to keep the data in a form other tools can read, and a vendor who locks it in has a problem. A useful test is to finish the sentence “if this is true, then we should…” If nothing comes after “should,” keep digging. Chapter 9 is about what gets built on top. The thesis has to survive being retold The second is that the thesis has to be short and plain enough for someone else to say it, accurately, in a room you are not in. That someone is the champion, as MEDDIC defines the role in Chapter 4: the person inside the buyer’s company who will sell for you when you are not there. Every large deal has meetings the seller never attends. The budget review. The executive committee. The conversation between the chief finan‐ cial officer and the chief information officer in a corridor after the meeting you were in. Gartner’s research on business buying found that buyers spend only about 17 percent of their buying time with suppliers, all suppliers combined. The decision is made mostly in the other 83 percent. What your champion carries into those rooms is whatever they can remember. A feature list does not survive the trip. By the time it reaches the budget review it has become “they said it was pretty good” and a slide nobody opens. A thesis survives if it is one sentence in plain words. Picture a chief data officer at an insurer in her own leadership meeting, a week after your call. Someone asks why the company should spend money on this now. She does not reach for your deck. She says, “Our definitions of a claim and a loss will outlive any tool we buy, so we should get them written down first, and buy something that keeps them outside the AI model.” The room can argue with that, and it is your argument, carried whole by someone else. So write the thesis for the person who will repeat it, not for the person who wrote it. If a tired executive cannot say it back after hearing it once, it is too long. 92

Page 95 of 214. What a Thesis Is, and How to Test One The thesis has to outlast the next release The third requirement is time. A thesis has to stay true longer than the product in front of you. In 2012, at Amazon’s re:Invent conference, Jeff Bezos was asked about the future. “I very frequently get the question: ‘What’s going to change in the next 10 years?’” he said. “And I almost never get the question: ‘What’s not going to change in the next 10 years?’ And I submit to you that that second question is actually the more important of the two, because you can build a business strategy around the things that are stable in time.” In retail, he went on, customers will always want low prices, fast delivery and a vast selection. Nobody, ten years from now, will ask Amazon to deliver a little more slowly. That is the test of a thesis’s durability, and it matters more in AI than almost anywhere. The Introduction described the gap between capability and adoption: the AI models change in ways that matter every few months, and enterprises take years to absorb what they can do. A thesis tied to what this quarter’s AI model can do is already expiring. “AI model X beats AI model Y on benchmark Z” will be false, or irrelevant, within a release or two, and a buyer who built a plan on it will have to rebuild the plan. The buyer’s own world moves much more slowly. Their data, their regulators, their customers, the pressure on their costs and the risk they can stand will be much the same in five years. Build the thesis on those. “Our data will outlive every AI model we run on it” was true before this year’s AI models and will be true after next year’s. For an enterprise seller this changes where the argument lives. The product claim can change every quarter, and it should, because the product will. The thesis holds while the product changes under it. That is also why durability belongs with the second test. A thesis that is true whoever wins the deal is usually true whichever AI model is current, because it is about the buyer’s world and not about anyone’s product. “ A thesis that can survive any outcome says nothing. 93

Page 96 of 214. Thesis-Based Selling Candidate statements, run through the tests Here are eight sentences of the kind that turn up in a sale, and what the tests make of them. “Our platform answers questions twice as fast as the market leader.” Fails the second test. It is only true if you win. It is a claim about a product, and Chapter 6 shows how to dig under claims like it. “AI will change everything.” Fails the first test, and the third. Nobody disagrees, and nothing could prove it wrong. It is a truism. “Innovate with confidence.” Fails all three. It is a slogan. “The solution must support single sign-on and keep our data in Europe.” Not a thesis at all. It is a requirement. It says what the buy‐ er needs, not how the world works, and nothing could prove it wrong because it is not a claim. “The market will come to its senses about AI.” Fails the third test. Which senses, by when, and what would we see? As written, no out‐ come could contradict it. “Within two years you will swap the AI model under your appli‐ cation at least once.” Passes. Someone could say no, it holds whoev‐ er sells the application, and it is wrong if two years pass without a swap. “Our data will outlive any tool we run on it.” Passes. A buyer who replaces their data every time they change tools would disagree. It is true whichever tool they pick, and it is wrong if the tools turn out to last longer than the data. “Every employee will have an AI.” Passes. It is the first of my five in Chapter 7, which also gives what would prove it wrong. A sentence that fails is not wasted. A claim can be dug under until a thesis appears. A slogan can be made specific enough to argue with. A requirement usually sits on a thesis that nobody has said out loud. Notice too that the three that pass are all about the buyer’s world, not about anyone’s product. That is the shape to aim for when you write your own. 94

Page 97 of 214. What a Thesis Is, and How to Test One Where a thesis comes from: three layers Sooner or later a seller asks where theses come from. From the product? From the people who built it? From the seller? From the buyer? The answer is that there are three layers, and the seller’s job is to work between them. The vendor’s thesis. Every product is a bet. The company that makes it believed something about the world and built the product on that belief, the way Tesla and SpaceX did in Chapter 2. The seller inherits that thesis on the first day of the job. It is real, and you should hold it honestly. But it is not the deal. A buyer who is asked to adopt your company’s thesis wholesale hears a pitch. The buyer’s theses. The buyer holds beliefs about their business and its future before any vendor arrives: about their data, their people, their regula‐ tors, their market. They are usually unspoken, and they do not depend on any product. Chapter 6 shows how to dig them out from under a decision. The shared thesis. The short list both sides agree on, from which the decision follows. It is built in the deal, found and written together. It is the only layer the sale stands on. THE VENDOR’S THESIS SHARED THESIS THE BUYER’S THESES the sale underneath both: market theses, neutral ground Figure 1 0 Three layers. The sale stands only where the two circles meet. So the seller maps between them. In round one, surface the buyer’s theses, and treat what you write as your best guess until the buyer confirms it. Hold your vendor’s thesis honestly, and know where it differs from theirs. In Agree, find the overlap and write it down with the buyer. In Build, make the case for the product from the shared thesis, not from the vendor’s. And 95

Page 98 of 214. Thesis-Based Selling when there is no overlap, when your product’s thesis runs against what the buyer believes, as it would have for a vendor selling Amazon a system with no interface in Chapter 2, there are two honest moves: have the thesis conversation, or walk away early. One more source is worth knowing. Some theses belong to neither side: claims about the market as a whole, like my five in Chapter 7. A thesis about where AI is going in every company is not your product and not the buyer’s plan, which makes it useful neutral ground for a first meeting. The layers also sort out a word the book has so far used for two things. Your theses are claims about the world that you carry from account to account, as in this chapter and Chapter 7. Their theses are the beliefs under one buyer’s decision, as in Chapter 6. The sale happens where the two meet. One deal, many theses: how each role sees it So far I have talked about the buyer’s theses as if the buyer were one person. In a large deal the buyer is a committee, and each person on it answers for something different. That shapes what they believe. Ideally a deal rests on one overarching thesis that everybody in the room shares. In practice, every role arrives with theses of its own, and in one sales cycle you will meet several. Here is a buying group I have met, in one form or another, on almost every enterprise AI deal. The people are a composite; the positions are the ones I hear. The chief data officer believes the data is the asset. Whatever we adopt has to make our data more valuable, not more scattered. She has spent years cleaning and governing it, and every new tool that copies it somewhere else undoes some of that work. The chief information security officer believes that nothing touches sensitive data without controls we can audit. He will not be persuaded by what an AI model can do. He will be persuaded by who can see what, and how he will prove it to an auditor. The chief financial officer believes that spend has to track measurable value. She has watched cloud bills grow faster than anyone planned, and an open-ended consumption contract for AI looks to her like the same mistake again. 96

Page 99 of 214. What a Thesis Is, and How to Test One The chief information officer or chief technology officer believes that we cannot be locked in. Whatever architecture they choose has to survive the next AI model, and the one after that. Most of them have lived through at least one painful migration, and they will not sign up for another. The line-of-business leader has the simplest thesis of all: my team’s throughput, this year. They need results now, not in three years. Procurement and legal believe that the risk has to sit somewhere we can see it, in terms they can enforce. And the people who will use the tool believe something nobody asks them about: this will either make my job better, or it will be one more thing to log into. None of these people is wrong. Each thesis is true from where that person sits. And some of them pull against each other. The chief data officer wants the data in one governed place; the security officer wants it touched by as few systems as possible; the line-of-business leader wants it in their team’s hands by Friday. Here is how I work with that. First, I map the roles and write down each one’s thesis, in their words where I can get them. Strategic Selling taught sellers to map the buying influences, and The Challenger Customer, the 2015 follow-up to The Challenger Sale by Brent Adamson, Matthew Dixon and colleagues, taught them to find the people inside the buyer who can move the group; this is the same map, with one more column: what each person believes. Second, I look for the overlap. Under all of the theses above sits one that every one of them can accept: our data and what we mean by it will outlive any AI model we run on it, so we will keep them under our control, in one place, and let the AI models come to them. That is the overarching shared thesis. The chief data officer hears her asset protected. The security officer hears control. The chief technology officer hears no lock-in. The finance chief hears that the spend goes into something that lasts. Third, I keep the role theses alive under the shared one, because each person still needs to hear their own. The line-of-business leader needs to hear what their team gets this year. The finance chief needs a measure. The thesis session with my engineer is where we find the conflicts before the customer does: if the security officer’s thesis and the line-of-business leader’s cannot both be met, we would rather know in our own room than in theirs. And the champion carries the overarching thesis into the rooms we are not in, while each role hears its own version from us. 97

Page 100 of 214. Thesis-Based Selling Walk back through what that does. One sentence everyone can stand on, and a short list of role theses that hang from it. Without the shared thesis, you are running five separate sales to five people who will compare notes. Without the role theses, the shared one sounds abstract to everyone. The skill is holding both. So the job for you, reading this, is to work out your own vantage point. Which roles are in your deal? What does each one answer for? What would each one say is true? You will often need several role theses in one sales cycle, all standing on one shared thesis. The Thesis Check in the Field Guide asks whether you know them. How many: three to five How many theses should a deal rest on? Three to five. A thesis is foundational. It is what the decision stands on, and a founda‐ tion has few parts. Write twenty statements and you no longer have theses. You have requirements, and twenty requirements is a specification, which is the document this book is trying to get you out of. Fewer is harder, and better. Cutting to three forces the question of what truly matters, and the answer is usually the most useful thing you learn about the deal. Three to five also fits where a thesis has to live: in the memory of a room, and in the retelling of a champion who has to say it in meetings you will never attend. If you have eight, some of them are still claims, or two of them are the same belief in different words. Ask why again. What a thesis is not A thesis has near neighbors, and a seller can slide into any of them without noticing. Each one looks like a thesis from a distance. Each one fails in the room. Here they are, with how a seller turns a thesis into each, and what it costs. A pitch. A pitch is a claim about your product, made to win the deal. Our platform keeps your definitions safe. The thesis it came from was our data will outlive any tool we run on it. The seller turned it into a pitch the easy way: by attaching the product name. It is the most nat‐ ural move in selling, because the seller is paid to advocate for the 98

Page 101 of 214. What a Thesis Is, and How to Test One product and has been taught to bring every conversation back to it. The cost is immediate. The buyer stops evaluating a claim about their world and starts evaluating a vendor, and discounts everything accordingly. The thesis would have survived a competitor’s win. The pitch dies with the deal. An opinion. An opinion is a claim nobody has agreed to defend. I think most of your agents will be built by other agents within a few years. Said that way, at the end of a meeting, it is a remark. It becomes a thesis when the seller writes it down, says what would prove it wrong, and invites the buyer to attack it. Sellers drop that last part because it feels risky: naming a falsifier looks like handing the buyer a weapon. The cost is that the buyer has nothing to test, so the claim carries no more weight than the seller’s person‐ ality. An opinion is easy to nod at and easy to forget. A hypothesis. A hypothesis is a guess about one account, made to find the right problem, and thrown away when discovery proves it wrong. Chapter 4 draws the line in full. The careless move here is the reverse of the others: treating a thesis as disposable, dropping it at the first objection the way you would drop a wrong guess. A thesis is meant to be defended, bounded and revised in the open. A seller who abandons it at the first push teaches the buyer that nothing they said was meant. A slogan. A slogan is a sentence designed to be agreed with. AI you can trust. Data without limits. It is usually marketing’s tagline, and sellers repeat it because it is approved, short and everywhere. A slogan fails the first test, because nobody disagrees, and the third, because nothing could prove it wrong. The cost is subtle. The room nods, the seller hears agreement, and nothing has been agreed. When the champion tries to retell it in the budget meeting, there is nothing to retell: “they said their AI is trustworthy” persuades nobody. A requirement. A requirement is a condition the buyer has set. It must keep our data in Europe. It is important, and it is not a belief about how the world works. Sellers mistake requirements for theses because they are the most concrete things a buyer says. The way through is to ask what the requirement rests on. Data must stay in Europe because the buyer believes the regulators will keep tightening, and that belief is the thesis. Build on the requirement alone and you will meet it and still lose, because the buyer’s real reason was one level down. 99

Page 102 of 214. Thesis-Based Selling A prediction. A prediction is a claim with a date on it. By next March, half your analysts will use an assistant every day. It is useful, and it expires on the date. A careless seller turns a thesis into a prediction by making it more specific than the evidence allows, usually to sound confident. When March comes and the number is forty percent, the whole foundation looks wrong, even though the thesis underneath, that every employee will have an AI, is still on course. A thesis has a horizon and a signal to watch. A run of predic‐ tions, kept and reread, is how you find your theses, which is why you dated your bets before you started reading. Why sellers slide, and what the best ones do instead All six slides have the same cause. Sellers are excited about their products, and they should be. They are paid on the product, measured on the product, and trained by the people who built it. Marketing hands them messages. The quota resets every quarter. Every pressure in the job pulls the conversation back up to the product, where the pitch, the slogan and the confident prediction live. The sellers who win trust go the other way. In my experience the best of them go two or three levels below the product, to the belief underneath it, and sell that. They are not selling a database. They are selling the belief that a company’s data will outlive its tools, and the database follows. They are not selling an assistant. They are selling the belief that every employee will have one, and that the question is what each one is allowed to see. That is also why the strongest sellers I have worked with carry their success from company to company. The product changes every time they change jobs. The skill does not. Finding the beliefs under a decision, saying them plainly and getting a room to agree on them works in any account, for any product, in any year. A seller who only knows the product starts again at every new company. A seller who knows how to sell a belief brings the whole craft along. 100

Page 103 of 214. What a Thesis Is, and How to Test One Why not call it an axiom I studied mathematics, and my instinct, faced with any hard question, is to go back to the axioms: the first principles everything else is built on. For a long time the working title of this book was The Axiomatic Approach to Selling Enterprise Software. So the word deserves an explanation of why it is not on the cover. An axiom is a statement accepted without proof, the starting point from which everything else in a system is derived. The arithmetic you learned at school rests on a handful of them. In 1889 the Italian mathematician Giuseppe Peano wrote them down for the natural numbers. He started counting from one; the modern form starts from zero, and in plain words it says: Zero is a number. Every number has a next number, its successor. Zero is not the successor of any number. Different numbers have different successors. If something is true of zero, and whenever it is true of a number it is also true of the next, then it is true of every number. Everything about counting follows from those five lines. How much follows, and how carefully, is a famous story. In Principia Mathematica, Alfred North Whitehead and Bertrand Russell set out to derive mathematics from logic alone. They reach the groundwork for 1 + 1 = 2 several hundred pages into the first volume, in 1910, and complete the proof in the second volume, in 1912, with a dry remark beside it: “The above proposition is occasionally useful.” The older example is Euclid’s, which Chapter 9 comes back to. His geometry rests on five postulates. For two thousand years mathemati‐ cians tried to prove the fifth, about parallel lines, from the other four, and failed. In the nineteenth century Nikolai Lobachevsky, János Bolyai and later Bernhard Riemann tried something else: they replaced it. Change that one foundation, and a whole new geometry follows, con‐ sistent in every step, describing a different world. Einstein’s general rela‐ tivity would later need one of them. 101

Page 104 of 214. Thesis-Based Selling That is the most useful idea in this book, and mathematics showed it first. Change the foundation, and a different world follows from it. A thesis works the same way in a sale. Change one belief the buyer holds, say from “AI is an experiment for the innovation team” to “every employee will have an AI,” and a different budget, a different architecture and a different short‐ list follow from it. The idea of a starting point taken as given reaches well beyond mathe‐ matics, which is part of why I nearly used it. The Declaration of Indepen‐ dence opens its second paragraph with the words “We hold these truths to be self-evident,” and then lists the starting points a new country would be built on. Those are axioms in all but name: statements the document does not try to prove, from which everything after them follows. So why not call this the axiomatic approach? Two reasons. The first is that the word loses a room. I love it. Most sellers do not. Say “axiom” in a sales meeting and half the table stops listening. Everybody understands what it means to have a thesis, and investors have already taught executives to respect the word. The second reason matters more. An axiom is too strong. Inside its system it is accepted without proof and never questioned. The foundation of a deal has to be the opposite: stated, examined, tested, and revised when the evidence turns. No buyer owes a vendor an axiom. A thesis asks to be argued with, and the arguing is what produces belief. People buy the thesis, and they come to hold one by testing it against what they already take to be true. Nobody adopts a belief because a seller told them to assume it. Before I leave the mathematician’s vocabulary, three more of its words are worth knowing, because a seller meets their meanings every day. A postu‐ late is an assumption you ask the other person to grant so the work can begin. A supposition is weaker, something assumed for the sake of argu‐ ment. A predicate is what you assert about a subject, and to say a decision is predicated on a belief is to say the decision rests on it. That last one names the heart of the method: in a sound sale the action is predicated on the thesis. If this is true, then that follows. The thesis is the right word for this book because it carries both halves of the job. It is a foundation, like an axiom, and it is put up for argument, like Luther’s ninety-five. It is set down to be examined, agreed and built on, and it stays open to being changed. 102

Page 105 of 214. What a Thesis Is, and How to Test One FIELD NOTE 06 A thesis is a claim someone in the room could say no to. If everyone would nod, keep sharpening it. The thesis checklist Put the three tests and the three demands of the room together, and you have six checks for any thesis before it goes in front of a customer. The work‐ book’s Write it down page carries them. The sentence 1 Someone in the room could say no to it. 2 It is true no matter who wins the deal: no product name, and it survives being read by a competitor. 3 It names what would prove it wrong, and that evidence is something a person could observe. In the room 4 Something follows from it: if the buyer agrees, a plan changes. 5 A tired executive can repeat it after one hearing: one sentence, plain words. 6 It has a horizon and a signal to watch, and does not depend on this quarter’s AI model. The first three checks are about the sentence itself, and you can run them at your desk. The last three are about what happens when the sentence meets a room, and they are the ones sellers skip. Run all six on one thesis you hold today. Most first drafts pass the first two, fail the third because nothing would prove them wrong, and fail the sixth because they lean on what this 103

Page 106 of 214. Thesis-Based Selling year’s AI model can do. Fix those two and the thesis is usually ready for the next chapter’s work. YO U R TAS KS What to do this week Write three to five. On paper, before you open any tool, write three to five sentences you believe about where AI is going in your customers’ industry. One sentence each, in plain words, as if you had to say it to a chief executive in the elevator. Run the six checks. Be hard on them. In my experience most first drafts fail the first test, because nobody would disagree, or the third, because nothing could prove them wrong. That is normal. It tells you where to sharpen. Strip the product out. Reread each sentence as if you worked for your strongest competitor. If it stops being true, it was a pitch; rewrite it until it would hold whoever wins. Write the falsifier. Under each thesis, finish the line I would stop believing this if… and make the answer something a person could actually observe, not a feeling. Keep one. If only one thesis survives all six checks, keep that one. One sentence you can defend is worth more than ten that fall over at the first question. 104

Page 107 of 214. Chapter Six Find the Beliefs Your Deal Rests On “The unexamined life is not worth living.” Socrates IF THIS IS TRUE Every decision already rests on a few beliefs, whether anyone has said them or not. THEN You find your theses by digging under a decision until you reach the beliefs that hold it up. 105

Page 108 of 214. Thesis-Based Selling The last chapter defined a thesis and gave you the tests for one. If we stopped there, you could recognize a thesis when someone handed you one. In a deal, nobody hands you one. The theses are buried under the decision, usually unspoken, often unnoticed by the buyer and the seller alike. This chapter shows you how to dig for them, step by step: list the reasons, ask why until you reach a belief, and write it down. Take a decision with no technology in it. A couple is deciding whether to spend six thousand dollars on a week at the beach. One of them has three browser tabs open and is comparing resorts. This one has a pool for the children. That one is four hundred dollars cheaper. The flights are a bargain if they book by Friday. The other keeps finding reasons to wait. Maybe next year. Maybe somewhere closer. They think they are disagreeing about resorts. They are disagreeing one level down. Under the first person’s tabs is a belief: we deserve to have fun, and it is okay to spend money if we have fun. Under the second person’s hesitation is another: money is safety, and spending it on pleasure is a risk. Neither has said it out loud. No discount on the room will settle it, because the room was never the question. Every decision works this way. Buying a database, leading a company through a change, booking a holiday: each one sits on a few beliefs about how the world is, whether we are conscious of them or not. Establishing your theses means going down to find them, saying them in plain words, and choosing which ones you are prepared to defend. The same is true inside a company. How it does security rests on a thesis. So does the way it buys software, the market it chooses to enter, how it treats its people, and the culture that grows out of all of that. Most of those theses were never written down. 106

Page 109 of 214. Find the Beliefs Your Deal Rests On Claims compare products; theses decide deals Most of what gets said in a sale is claims. “The product is fast.” “The product is cheaper.” “The product is secure.” “The team already knows it.” A claim is about the product or the situation in front of you, and it can be checked against a competitor’s claim. That is why sellers like claims: they fit on a slide and they can be won. A thesis is the belief that makes a claim matter. “The flights are cheap” is a claim. The cheap flights only matter to someone who already believes the trip is worth taking. To the partner who believes money is safety, a cheap flight is still money leaving the account. In a deal it works the same way: “our product is cheaper” only lands with a buyer who believes cost is the thing to optimize, and it falls flat in front of a buyer whose real belief is that the last cheap choice cost them a year. Chapter 5 gave six checks for a well-formed thesis: three tests and three demands of the room. Those check the sentence once you have it. They do not tell you where to find it. Asking why does that job: it tells you whether you have reached the foundation or stopped at a claim on the way down. Ask “why does that matter?” and keep asking. Toyota taught its engi‐ neers to ask why five times to get from a symptom to a cause, a practice Sakichi Toyoda began and Taiichi Ohno made part of the Toyota Produc‐ tion System, and the same habit works here. A claim gives way to another claim underneath it. A thesis gives way to nothing. When the only answer left is “because that is how the world is,” you have arrived. Another way to tell the two apart is to take the statement away and look at the decision. If the decision still stands, the statement was a claim. If the decision falls over, the statement was carrying the weight. “We will not run our own data centers” decides the database, the ware‐ house and the next ten purchases after them. If a statement only ever decides one thing, keep asking why. 107

Page 110 of 214. Thesis-Based Selling The flights are cheap A CLAIM why does that matter? The trip is worth taking A CLAIM why does that matter? We deserve to have fun A THESIS why? because that is how the world is Figure 1 1 Keep asking until the answer stops being a claim: that is the thesis. The ladder in the figure is the method in one picture. Start from the claim people actually say: the flights are cheap. Ask why that matters, and you get another claim: the trip is worth taking. Ask again, and the answer stops being about the trip: we deserve to have fun. Ask why once more, and the only answer left is “because that is how the world is.” That is how you know you have reached a thesis. In a deal the rungs look different, but the climb is the same, and most sellers stop on the first rung. FIELD NOTE 07 A claim settles one comparison. A thesis settles the next ten decisions. From a decision to three to five theses The procedure has three steps, and they are round one of the workbook at the back of this book. 1 List. Write the decision your customer is facing at the top of a page, in their words and without a product in it. Under it, list every reason you have heard for it, and every reason you have heard for waiting. Ten to fifteen in all is normal. Leave them unsorted. The reasons to wait matter as much as the reasons to act: at the beach, the decisive belief was under the hesitation. 108

Page 111 of 214. Find the Beliefs Your Deal Rests On 2 Ask why. Under each reason, ask why it matters until you reach a belief about the world. Fifteen reasons usually land on three to five beliefs. Several reasons that looked unrelated turn out to share one root. 3 Write it down. Write each belief as one sentence. Run the six checks from Chapter 5, and write what would prove each one wrong. Then put them in order, so that each stands on the one before. Aim for three to five, for the reasons Chapter 5 gives: enough to build on, few enough to remember and retell. If you have eight, some of them are still claims, or two of them are the same belief in different words. One caution about what round one gives you. You did it on your own, about the customer’s decision, so the theses you wrote are your best guess at what the buyer believes. Treat the page as a draft. It becomes theirs only when they have heard it and responded, which is the first step of round two. Three worked examples Here are the same three steps, run on three decisions of very different size. 109

Page 112 of 214. Thesis-Based Selling The decision Spend six thousand dollars on a week at the beach this summer. What gets The flights are cheap right now. The resort has a pool for the children. We have not been away in two years. said Ask why Cheap flights matter only if we are going. The pool matters only if the trip is worth taking. Two years without a break matters only if a break is good for us. The theses underneath 1. We deserve to have fun. 2. It is okay to spend money if we have fun. 3. The children will be this age once. What follows Go this summer. Choose the place for the time together, and stop comparing room rates. Figure 1 2 Worked example 1: A week at the beach Walk back through what the worksheet did. The first row put the deci‐ sion in one sentence, with a price on it. The second wrote down what actu‐ ally gets said in a conversation like this: cheap flights, a pool, two years without a break. Those are claims, and each one sounds like a reason. The third row asked why each claim matters, and every answer came back condi‐ tional: the flights matter only if we are going, the pool only if the trip is worth taking. Each claim was leaning on something underneath it. The last row is what follows once the belief under all three is said out loud: if a break together is good for us, then go, choose the place for the time together, and stop comparing room rates. I chose a beach holiday because it has no technology in it, so nothing distracts from the shape. And the shape is the whole lesson. The claims compared resorts. The decision was never about resorts. Hand the same brochure to someone whose thesis is that money is safety, and none of it lands. The pool, the price and the flights are answers to a question that person has not agreed to ask. Three takeaways. Claims are what people say; theses are why the claims matter. A disagreement about options is often a disagreement one level down. And until the belief underneath is agreed, more information about the options only makes the argument longer. That is round one on a small 110

Page 113 of 214. Find the Beliefs Your Deal Rests On scale, and it is exactly what happens in a deal when the buyer’s team argues about vendors and is really arguing about what they believe. The decision Replace the database behind the company’s reporting. What gets It is faster on our queries. It costs less per year. Our engineers already know it. It reads open formats. said Ask why Speed matters only if we stay long enough to care. Cost per year matters only if the workload holds steady. Our engineers knowing it matters only if we keep running it ourselves. Open formats matter only if we expect to change engines one day. The theses underneath 1. Our data will outlive every tool we run on it. 2. We will not run our own machines. 3. We cannot predict next year’s workload. What follows Open storage first, a managed service, and pricing that follows use. Then compare speed among the products that fit. Figure 1 3 Worked example 2: Buying a database Now the same steps on a decision a seller meets every quarter. The team says four things about the new database: it is faster on our queries, it costs less per year, our engineers already know it, and it reads open formats. Ask why each one matters and four conditions appear. Speed matters only if the company stays on this engine long enough to care. Cost per year matters only if the workload holds steady. The engineers’ familiarity matters only if the company keeps running the database itself. Open formats matter only if the company expects to change engines one day. Under those conditions sit the three theses in the table: the company’s data will outlive every tool it runs on it, the company will not run its own machines, and nobody can predict next year’s workload. Each one turns a claim into a direction: open storage first, a managed service rather than a team running servers, and pricing that follows use. Look at what happened to speed. It was the first thing said and it is the last thing decided. Once the three theses are on the table, speed gets compared only among products that already fit them. That is the lesson of 111

Page 114 of 214. Thesis-Based Selling this example for a seller. If your product is the fastest but locks the data into its own format, you lose to a slower product that keeps the data open, and no benchmark will save you, because the buyer is no longer asking which is fastest. They are asking which ones fit what they believe. So in your own deals, when a buyer leads with a feature, ask why it matters before you answer it. The feature is usually a claim, and the belief under it decides which features will count. Find the belief in round one, and you know which comparison you are really in before your competitor does. The decision Roll out AI to every team in a company of five thousand people. What gets Our competitors are doing it. The board keeps asking. It will save hours. The tools are finally good enough. said Ask why "Competitors are doing it" is a fear, and it points nowhere. "It will save hours" matters only if the work keeps its present shape. Keep asking, and the question becomes what work looks like once these tools are everywhere. The theses underneath 1. Every employee will have an AI. 2. Every employee will have more than one agent working for them. 3. What you give AI is worth more than what AI gives you. What follows Teach everyone to delegate and to state intent. Decide where the rules for agents live. Measure the work that changed, and stop counting hours saved. Figure 1 4 Worked example 3: Leading a company through AI The third example is the largest decision of the three, and the one most of your buyers are facing now. Notice the claims the leadership team makes: competitors are doing it, the board keeps asking, it will save hours, the tools are finally good enough. The first is a fear, and a fear points nowhere: knowing that competitors are moving does not tell you which way to move. The third assumes the work keeps its present shape, which is exactly what AI is about to change. Keep asking why and the question stops being about tools at all. It becomes a question about what work looks like once these tools are everywhere. 112

Page 115 of 214. Find the Beliefs Your Deal Rests On What follows is a different plan from the one the company started with. Instead of counting hours saved, teach everyone to delegate and to state intent. Decide where the rules for the agents live. Measure the work that changed. Those three moves rest on theses about AI inside companies. Those are three of my own five, and Chapter 7 gives all of them. The takeaway for a seller is the scale of what a thesis can do. A seller who brings those theses to that company’s leadership, and gets them agreed, has handed five thousand people a way to work out their own next move, which is the starlings again. No feature list does that. Run the same three steps on your largest account this week, and see whether the plan you find under‐ neath matches the plan you were selling to. “ A thesis is the belief that makes a claim matter. Every buyer holds theses, named or not The beliefs are there whether you dig for them or not. A buyer who has never written down a thesis still has several, and they are deciding the purchase. A seller who has never written one down is still selling on one, usually the one the product was built on, and usually without knowing whether the buyer shares it. Digging does not create those beliefs. It lets you look at them, on paper, before they decide the deal for you. A belief you can see, you can check with the buyer, sharpen, or build on. A belief you cannot see just quietly votes against you in a meeting you are not in. Digging also shows you that there are two kinds of thesis in every deal. Yours are claims about the world that you carry from account to account, like my five in the next chapter. Theirs are the beliefs under this buyer’s deci‐ sion, which is what this chapter digs for. In Chapter 5’s terms, theirs are the buyer’s layer; yours are claims about the market that you bring as neutral ground, separate from your company’s product thesis. Where the two match, the rest of the sale is short. Where they conflict, the conversation has to be about the thesis, and if it cannot move, the honest thing is to say so early. 113

Page 116 of 214. Thesis-Based Selling A belief you have said out loud can be tested, agreed with, argued against and changed. One you have never said can only be obeyed. YO U R TAS KS What to do this week Start small. Run the three steps on a personal decision you made this year: a trip, a car, a move. It is easier to see the beliefs under a decision when no quota depends on it, and you will find two or three you had never said out loud. Then a deal. Run the same steps on one live deal. Write the decision in the customer’s words, list the reasons for it and the reasons for waiting, ask why under each until you reach a belief, and write the beliefs down. Cut to three to five. Ask why again under each belief, and drop what turns out to be a claim about a product. Merge the ones that are the same belief in different words. What is left is your draft. Order them. Arrange them so that each stands on the one before, and mark the page as a draft. It becomes the customer’s only when they have heard it and responded. 114

Page 117 of 214. Chapter Seven Five Theses About AI Inside Companies “What important truth do very few people agree with you on?” Peter Thiel IF THIS IS TRUE A set of theses is strongest when each one stands on the one before. THEN Five short claims can carry a buyer from a seat license to the question of who governs a company’s agents. 115

Page 118 of 214. Thesis-Based Selling The last chapter dug under decisions by asking why, and then asking again. Any parent knows where that ends. A child asks why the sky is blue, then why the light scatters, then why that is, and somewhere around the fifth why the parent runs out of answers and says, “Because that is how it is.” That moment is not a failure. It is the ground. Toyota’s five whys, from the last chapter, were built to reach exactly that point. In a sale, the ground you reach is a thesis, and everything else is built up from it. So here are mine. These are five theses I hold about AI inside companies, as of this edition of the book. You may disagree with some of them, and that is fine. It is even the point. This chapter is not here to win the argument about whether they are right; time will tell how well they age, and I expect some of them to need a boundary or a revision before long. It is here to show what a full set of theses looks like when it is written the way this book asks, and how an enterprise seller builds on a thesis through a deal. Hold them up against the five dated bets you wrote before you started reading. Where yours differ from mine, you have found something worth arguing about. I almost left my own theses out of this book. They will date, and a book that argues for writing theses down should not pretend its author’s are permanent. I kept them because the method is easier to learn on a real set than on an abstract one. Read them as a worked foundation. The chapters that follow put them to work: Chapter 8 takes the fifth into a thesis session, Chapter 9 builds a sale on it layer by layer, Chapter 10 defends the first against three kinds of pushback, and Chapter 11 attacks the fifth with a machine. Here they are at a glance, before I take them one at a time. 1 Every employee will have an AI. 2 Every employee will have more than one AI agent working for them. 3 AI’s capacity for work will be greater than human capacity for work. 4 More AI agents will be created by other agents than by humans. 116

Page 119 of 214. Five Theses About AI Inside Companies 5 What you give AI is worth more than what AI gives you. For each one I give the claim, why I hold it, what would prove it wrong, what follows if it holds, and what it means in a deal: the decisions the buyer faces, the options they have, who cares, what to ask, and how it shapes the case for a product. They are in order, and each stands on the one before. Under each is a short scene, an illustration of a seller using the thesis with the executive who has to lead the change. The chapters after this one use the five as their running example. Several of the theses that follow rest on one word, agent, and it is a word everyone uses differently. So before I state them, here is exactly what I mean by it. What I mean by an agent The word is used loosely, and I would rather say so than pretend it has one meaning. One person sets up a tool that sends out their email campaigns and calls it an email agent. Another has something that reads their Slack channels each morning and writes a summary, and calls that a Slack agent. A research lab would call neither of those an agent. For this book the definition is practical. An agent is software that has been given a job, has access to what the job needs, and does the work without a person performing each step. The email sender and the Slack reader both count. So does a system that replans a supply chain overnight. The definition changes the numbers in what follows and leaves the argu‐ ment where it is. Count every small helper and the crossover in the third thesis comes sooner. Count only the ones that act on their own judgment and it comes later. Either way the line is rising. With that word settled, here are the five, in the order they build. 117

Page 120 of 214. Thesis-Based Selling 1 THESIS 1 OF 5 Every employee will have an AI. Why I hold it. Every general tool of knowledge work started with a few specialists and ended on every desk: the telephone, the personal computer, email, the smartphone. AI assistants are on the same path and moving faster, because they arrive inside software people already use and are sold by the seat like the software around them. It is wrong if companies run the trials, find that an assistant does not pay for its seat, and take it away from most of their staff. Then an AI becomes standard issue, like a laptop. The useful question stops being whether to give people one and becomes what each one is allowed to see and do. In a deal. If this holds, the buyer’s decision is not whether to buy assistants. It is how to govern them. The options are assistants built into software the company already runs, a standalone assistant platform, or building on an AI model provider directly, and each puts the control in a different place. The chief information officer cares about the standard, the chief infor‐ mation security officer about what each assistant can see, human resources about who gets one and how they learn it, and finance about the cost at full scale, not at forty seats. The discovery question is simple: “If every one of your people had an assistant by next year, what should each one be allowed to see?” The shared thesis to aim for is that assistants will be universal, so access and permissions are the real decision. A product case built on that rests on control of data, not on this month’s features. Picture it. A sales engineer is meeting the chief information officer of a regional bank about a trial for forty compliance analysts. She mentions that every division head has since asked for the same thing. He stops talking about the trial. “If everyone will have one within the year, the forty seats are the small question. The large one is what each assistant is allowed to see.” The rest of the meeting is about that. 118

Page 121 of 214. Five Theses About AI Inside Companies Notice what she did not do. She did not argue features or price for forty seats. She put the first thesis on the table in one sentence, the buyer agreed with it out loud, and the deal grew from forty seats to a decision about access for the whole bank. That is the pattern every scene in this chapter follows: no reference customer for the bigger decision, so agree the thesis first and build the deal on it. 2 THESIS 2 OF 5 Every employee will have more than one AI agent working for them. Why I hold it. An assistant answers when asked. An agent is given a task and goes away to do it. Work is made of many tasks, and a person who has learned to hand off one will hand off ten: one agent watching the inbox, one preparing the weekly report, one checking the numbers before they go out. It is wrong if a single general agent turns out to handle everything, so the count stays at one. Even then the next thesis would stand, because the capacity would keep growing. Then every employee becomes a manager. Delegating clearly and checking the work, skills that used to belong to supervisors, become part of every job. In a deal. If this holds, the buyer is not buying a tool. They are changing every job into a small management job, and that raises decisions the purchase alone does not settle: who teaches people to hand off work and check it, how agent work is monitored, and whether agents come inside the applications people already use or from a platform that spans them. Line-of-business leaders care because their teams’ work changes shape; learning and development cares because delegation is a skill; IT operations cares because someone has to watch the agents run. Ask in discovery: “Who in your company is teaching people to hand work to an agent and check what comes back?” 119

Page 122 of 214. Thesis-Based Selling The shared thesis is that every employee becomes a manager of agents, and the product case follows from it: management, monitoring and training, with features last. Picture it. At a medical device company, an account executive asks the vice president of transformation how the complaint-handling rollout went. She says each investigator now works with three agents: one pulls the device history, one drafts the report, one checks it against the regulation. Nobody was hired and nobody was let go. The seller asks who taught the investigators to manage. Nobody did, and that becomes the next project, with the seller helping to shape it. The second thesis turned a finished rollout into a new conversation about training and monitoring, which the buyer had not known to ask for. 3 THESIS 3 OF 5 AI’s capacity for work will be greater than human capacity for work. Why I hold it. The arithmetic follows from the first two theses. A company of a thousand people where each has three agents has three thousand agents, and an agent can take the next task the moment it finishes the last. The count is the smaller part. An enterprise runs on capacity, the amount of work it can get done, and the intelligence of any single AI model does not supply that. A company needs intelligence and capacity, and capacity is the one it has never had enough of. It is wrong if agents stay narrow and unreliable enough that a person has to check every step, so that they add to the count without adding to the work a company can get done. Then people stop competing with AI on capacity and start steering it. The job is putting that capacity to work: deciding what the work is, handing it over clearly, and having the next piece ready, because capacity with no work in front of it is worth nothing. The org chart, the budget and the way work is assigned and reviewed were built for a workforce of people, and they have to be looked at again 120

Page 123 of 214. Five Theses About AI Inside Companies before the crossover arrives. A company should know roughly when its own crossover comes, and be prepared to put that capacity to work. In a deal. If this holds, the scarce thing is no longer capacity. It is work that is ready for the capacity. The buyer’s decisions become what work to point it at, how to redesign processes so that work flows to agents, and how to measure capacity used rather than hours saved. The options range from orchestration and workflow tools to observability for agents (tools that show what each agent is doing and how busy it is) to help redesigning the processes themselves. The chief operating officer and the chief financial officer care most, along with any transformation office. Ask: “How many agents run in your company today, and how much of what they could do is anyone asking of them?” The shared thesis is that capacity is outrunning use. A product case on that ground is about putting capacity to work, and it beats a case built on doing one task slightly faster. Picture it. At a telecom company with twenty thousand employees, a sales engineer asks the chief executive two questions. The first: how many agents are running across the company today? Nobody in the room knows. A week later the answer comes back, and it is thirty-one thousand. The second: how much of what they could do is anyone asking of them? Most were set up for one small job, which they finish in minutes, and then they wait. The company has more capacity than it has ever had, and nobody’s job is to keep it busy. The two questions did more than a demo could have. 121

Page 124 of 214. CAPACITY TO DO WORK Thesis-Based Selling what the agents can do what the staff can do from here on, people steer it THE CROSSOVER TIME Figure 1 5 The third thesis, drawn without a scale. Past the crossover people steer it. The figure has no numbers on purpose. Nobody knows the year the lines cross for any given company, and a chart that pretended to would be a prediction, not a thesis. What the figure shows is the shape: the capacity of the staff grows slowly, the capacity of the agents grows fast, and past the crossover the people’s job changes from doing the work to steering it. For a seller, the useful question is not when the lines cross in general. It is when they cross for this customer, and what that customer will have ready for the capacity when they do. 4 THESIS 4 OF 5 More AI agents will be created by other agents than by humans. Why I hold it. A person might set up five agents. To get their work done, those five may need to create and manage fifty more. Delegation compounds, and after the first layer no human is doing the creating. It is wrong if companies forbid agents from creating agents and the rule holds. It would cost them the speed that made agents worth having. Then something new exists inside the company: a population of agents that no person set up one by one, working at a scale and speed no person can supervise directly. I think of it as a civilization that sits outside the part 122

Page 125 of 214. Five Theses About AI Inside Companies and built into the place where the agents live, because nobody will be standing over each one. In a deal. If this holds, nobody can approve every agent, so the buyer has to decide where the rules live and who can change them. The decisions are about limits on what agents may create, identity for agents, audit trails, and a written page of rules, values and goals that every agent reads. The options are policy and governance layers, identity and access for non-human workers, and audit and logging, whether bought or built. The chief information security officer, risk and compliance, legal and the chief technology officer all have a stake, and they rarely sit in the same meeting. Ask: “Where do the rules your agents obey live today, and who changes them?” The shared thesis is that control has to be written into the environment the agents work in, and the product case becomes one about policy, identity and audit. Picture it. A technical consultant is with the chief operating officer of a manufacturer on the morning a port closes. By the afternoon the five agents her planning team set up have created more than fifty others, one for every affected part number, each working out a new plan. No person approved any of them. What kept the afternoon safe was a page the consultant had pushed her leadership team to write months earlier: use only qualified suppliers, and never commit money above a set limit without a person. This is the murmuration from the Introduction, built out of software: thousands of individual decisions, a handful of local rules, one coherent system. It is the only kind of control that scales to agents making agents. There is a name for what the flock does: emergence. No bird decides the shape, and no rule describes it. The shape comes out of many birds each following those few rules. A company full of agents will behave the same way. Its results will come out of what thousands of agents do next to each other, and nobody will have dictated them. That can go well or badly, which is why the few rules, values and goals you write down matter more than any single instruction. 123

Page 126 of 214. Thesis-Based Selling 5 THESIS 5 OF 5 What you give AI is worth more than what AI gives you. Why I hold it. Give a prompt for an app to an AI model from a few years ago and you get an acceptable app. Give the same prompt to a current frontier AI model, one of the most capable AI models the leading labs have released, and you get a much better one. Give it to next year’s AI model and the app will be better again. The output keeps being replaced. What held its value through all three was the thing the person asked for. It is wrong if AI models become able to work out what a company wants without being told, or if what they produce stays expensive to produce a second time. Then the asset to keep is the intent. Today companies save prompts and the artifacts that came out of them. They should be saving what they meant: the intent, the values, the goals, the mission. People become architects of intent more than makers of code and documents, and that is the work to hire and train for. Intent. We already treat intent as valuable everywhere else. The law judges the same act differently by what the person meant. A contract is read for what the parties intended. A manager who knows what the company is trying to do needs far fewer instructions than one who does not. Companies spent twenty years on data engineering: moving, cleaning and governing what they know. The next discipline is intent engineering: writing down what the company wants, its values, its goals and its limits, pre‐ cisely enough for a machine to act on, and keeping that page current. The data says what is. The intent says what it is for. In a deal. If this holds, the buyer’s most valuable AI asset is not any output and not any one AI model. It is what they give the AI models: the prompts, the definitions, the intent. The decisions are what to keep, where to keep it so every better AI model can use it, and how to avoid being locked to one AI model provider. The options are tools that manage prompts and intent, platforms that give access to many AI models, and semantic layers, 124

Page 127 of 214. Five Theses About AI Inside Companies which hold the company’s business definitions, what counts as a customer or a loss, outside any one tool. The chief architect and the chief data officer care, and so does procure‐ ment, which worries about lock-in. Ask: “When a better AI model arrives, what in your plan would have to be rebuilt?” The shared thesis is that intent is the asset, and the product case rests on portability: keeping what the company means outside any single AI model. Picture it. A sales engineer working with the chief architect of a pharmaceutical company persuades her to stop filing away prompts and outputs. Each group writes one page instead: what a good study summary must do, who reads it, what it must never leave out, which decisions it has to support. The first AI model given the page produces passable summaries. The next year’s AI model, given the same page, produces good ones. The summaries were thrown away and made again. The page was kept, and it is worth more now than when it was written. For the seller, the turn came when the architect stopped asking which AI model was best this quarter. Once she agreed that intent is the asset, the question became where to keep the page so every future AI model could use it, and that is a question a seller can build a whole product case on. How the five build on each other Read them in order and each is the “then” of the one before. 125

Page 128 of 214. Thesis-Based Selling 1 Every employee has an AI THEN 2 Each employee has several agents THEN 3 AI has more capacity for work than humans THEN 4 Agents create more agents than humans do THEN 5 Humans supply the intent, the part worth keeping Figure 1 6 Five theses, each one the consequence of the one above it. If every employee has an AI, they will soon have several. If they have several, AI’s capacity for work will come to be greater than human capacity for work. To put that much capacity to work, agents will hand work to other agents, so more agents will be created by other agents than by humans. And if people are no longer making or watching each agent, what people contribute is intent. That gives a seller a line of argument that starts far from any product. Picture a buyer choosing an AI assistant by comparing this month’s features. Ask a different question: THE QUESTION “If your agents will have more capacity to do work than your people within a few years, and most of them will be created by other agents, where do the rules they all obey live?” A buyer who accepts the theses is now evaluating something else, and the feature list from the first meeting has moved down the page. 126

Page 129 of 214. Five Theses About AI Inside Companies When you change direction, your agents do not hear it One consequence of the five has had very little attention from the people it will affect most. Ask how long it takes for a decision made by the chief executive to reach everyone in the company. In most organizations the route is well worn. There is an all-hands meeting. Managers brief their teams. The strategy deck goes round. Within a few weeks most people have heard that the direction changed, and the rest pick it up from the people next to them. The chain is slow and imperfect, and it exists. Now ask the same question about the agents. The telecom chief execu‐ tive from the third thesis has thirty-one thousand of them. When she changes the company’s priorities on a Monday, none of them is at the allhands. No manager briefs them. They go on doing what they were set up to do the month before, in good order and in the wrong direction. The larger part of her workforce is walking blind. The starlings do not have this problem, because each bird can see its neighbors turn. An agent sees only what it was given. This is where the fourth and fifth theses meet. If agents outnumber people and mostly create each other, nobody can go round and tell each one. The only way to turn them is to change the thing they all read: the written rules, values, goals and mission from the fourth thesis, which is the intent from the fifth. That page is the chain of communication. A company that keeps it current can change direction in a day. A company without one cannot steer its own agents at all. 127

Page 130 of 214. Thesis-Based Selling CHIEF EXECUTIVE THE PAGE rules, values, goals, mission MANAGERS TEAMS each one tells the next every agent reads the same page PEOPLE: A FEW WEEKS AGENTS: A DAY Figure 1 7 People hear a change through the chain. Agents read it on one page. On the left, the old chain: a chief executive tells the managers, the managers tell their teams, and a change of direction reaches everyone in a few weeks. On the right, the agents: none of them is in that chain, so the only way to turn them is to change the one page they all read. That page is the written rules, values, goals and mission from the fourth thesis. A company that keeps it current turns its agents in a day. A company without one cannot turn them at all. For a seller, that page is also a better first meeting than any demo, because it is a problem the buyer has and has not named. It also gives the seller a second question for that chief executive. THE QUESTION “When you change direction, how long until your agents know?” How to use the five in a meeting Five is too many for one meeting. A buyer can hold two or three new ideas in an hour, not five. So pick the two or three that bear on the decision in front of this buyer, and lay them down first. A bank deciding on forty assistant seats needs the first thesis. A manufacturer whose planning team is already 128

Page 131 of 214. Five Theses About AI Inside Companies running agents needs the fourth. The others can wait for the next meeting, or for the champion to ask. Expect disagreement. A buyer who nods at all five is being polite, and politeness is not agreement. The useful meeting is the one where two theses hold and one gets a boundary, because the boundary tells you exactly where this buyer’s world differs from the general case, and that is where the deal will be decided. And write your own. These five are mine. They are here to show the form, and to be attacked. Borrowed theses do not hold up: the buyer asks why, and you only know the answer I gave. The theses that carry a deal are the ones you have argued for yourself, against people who disagreed. “ A thesis you copied folds at the second question. If your theses differ from mine Suppose you hold a different set. Say your first thesis is the opposite of mine: most employees will never use AI directly; it will work in the background of the systems they already use. That is a real thesis. It passes the tests in Chapter 5: I would disagree with it, it holds whoever wins the deal, and it is wrong if, within two years, most knowledge workers are using an assistant every day. The method does not care which of us is right. Run the same six checks. Take it into a thesis session with the person who sells beside you. Lay it down in the first meeting and ask the buyer which parts they do not buy. Build on what they agree to, and defend it when they push. It will lead to different decisions. A buyer who agrees with your thesis invests in automating processes behind the scenes, not in seats. The people who care are in operations and IT more than in human resources. The question for discovery becomes which processes run without a person touching them, and the product case changes com‐ pletely. What stays the same is the shape: a thesis set down, agreed, built on and defended. Use mine as a worked example, then replace them with yours. That is the end of round one. You know what a thesis is, how to find the ones under a decision, and what a set of them looks like. Part Three takes 129

Page 132 of 214. Thesis-Based Selling them into the room, starting with the hardest room of all: your own team. YO U R TAS KS What to do this week Attack two. Pick the two of my theses you find weakest and write the best case against each, as if you were the buyer who disagrees. If you convince yourself, write the boundary that would make the thesis true again. Find the crossover. Estimate when the third thesis arrives at your company, or at your customer’s: count the people, guess the agents per person, and show the arithmetic. Then list the work you would hand that capacity on the first day, because capacity with nothing ready for it is wasted. Write one of yours. Write one thesis about AI inside companies that nobody else you know is carrying, and run it through the six checks. The set is unfinished until it holds one that is yours. 130

Page 133 of 214. Part III Sell on the Thesis Round two in practice: agree on the thesis, build the sale on it, defend it, and sharpen it before a customer tests it.

Page 134 of 214.

Page 135 of 214. Chapter Eight Agree with Your Own Team Before the Customer “Use the phrase ‘disagree and commit.’ This phrase will save a lot of time.” Jeff Bezos, letter to shareholders, 2 0 1 6 IF THIS IS TRUE The seller and the sales engineer walk into the meeting with two different pictures of what is true. THEN The first room to win is your own, before the customer hears a word. 133

Page 136 of 214. Thesis-Based Selling Part Two ended with round one done. You know what a thesis is and how to test one, you have dug the beliefs out from under a decision, and the last chapter laid out a full set of five as a worked example. If the book stopped there, you would already sell better than most, because you would know what you believe and why. But a thesis on your desk carries no deal. Establishing a thesis that passes the tests is the starting point, not the finish. Round one you can do alone. Then the hard work begins: getting agreement. First inside your own account team, between the account executive and the sales engineer. Then with the customer. In my experience this is the most delicate part of the whole method, because it is where the thesis meets other people, with their own instincts, incentives and fears, and a lot can go wrong. This chapter gives you the approaches that work, the conversations to have and the exact phrases to use, starting with the room most sellers skip: their own. Maya is an account executive. Daniel is the sales engineer on her deals. They are forty minutes into a call with Priya, the chief data officer of a regional insurer, and Priya asks the question she came to ask. THE QUESTION “Can this answer questions about our claims data and get them right?” Maya answers first. “Yes. Customers are seeing very strong accuracy.” Daniel waits a beat and adds, “It depends on how the claims data is modeled, and on what you count as right.” Priya writes something down. The call goes on, friendly, and ends on time. Afterward Maya is annoyed that Daniel undercut her, and Daniel is annoyed that Maya promised something he will have to deliver. Priya tells her team that the vendor does not seem to know what its own product does. Nobody lied. Maya answered from her picture of the truth, which is made of outcomes. Daniel answered from his, which is made of conditions. 134

Page 137 of 214. Agree with Your Own Team Before the Customer In a mature market the two pictures would have been pulled together by a shared stock of evidence: the last twenty deployments, which both of them had seen. Without that stock, each one fell back on the instinct of the role. Closing that gap is the first step of round two, Agree, and it starts inside your own team. Why the seller and the engineer disagree I have sat in Daniel’s chair. I was a sales engineer, and later I led teams of them. The scene is invented, and anyone who has done the job will recognize it. The account executive is paid when the customer commits, so the account executive speaks in results and dates. The sales engineer is the one who has to make it work on the customer’s data, so the sales engineer speaks in conditions and limits. A company needs both instincts. A customer needs to hear them as one voice. When they come out as two, the customer does a quick calculation. If the people selling this cannot agree on what it does, the risk of buying it is higher than the slide says. The JOLT Effect research from Chapter 4 found that fear of a mistake is what stalls most deals, and a visible gap between seller and engineer feeds that fear directly. The usual fix is a messaging document. Marketing writes the approved answer, and both people memorize it. That holds for one question. On the second follow-up the memorized line runs out, and each person goes back to instinct. The thesis session: agree before the meeting 135

Page 138 of 214. Thesis-Based Selling The fix that holds is to argue it out beforehand. Before the account plan and before the deck, the seller and the engineer sit down for forty-five minutes with one page. I call it a thesis session, and it has three steps. 1 Write separately, then compare. Each person writes three things they believe are true about this market. No products. Five minutes, no talking. Then put the two lists side by side. Where they match, you have a candidate. Where they differ, you have found the thing the customer would have found for you. 2 Attack each other’s list. The engineer attacks the seller’s list. Is this true on real data? Under what conditions? What would break it? Then the seller attacks the engineer’s list. Suppose it is true. Why would a chief financial officer care? What does it change? 3 Keep what survives, and write down what you do not know. A thesis has to hold up on real data and matter to the business. Three to five will make it through. Agree on the exact words. Then list the questions where the honest answer is “we have not seen enough to say.” Agree that both of you will say exactly that. The page that comes out is short: three to five theses, the evidence for each, what would prove each wrong, and the list of unknowns. Both names go at the bottom. 136

Page 139 of 214. Agree with Your Own Team Before the Customer BUYER hears one voice THE THESES one page, two signatures SELLER ENGINEER speaks in outcomes speaks in conditions Figure 1 8 Two instincts go into the page. One voice comes out of it. The figure shows what the thesis session does to the two instincts. The seller arrives speaking in outcomes, the engineer in conditions, and both are right. They go into the same page, get argued over, and come out as theses both people have signed. The buyer, on the other side, hears one voice. That is the whole point of the session: the gap between seller and engineer is going to be found, and it is far cheaper for you to find it in a forty-five-minute meeting than for the buyer to find it in a sales call. Why an agreed truth beats a message A message is something you were given. A truth you argued for is something you own, and it behaves differently under pressure. When Priya pushes for the fifth time, Maya has no script to run out of. She has the argument she already had with Daniel, who pushed harder than Priya ever will. Think about what happens on the call without the session. Priya asks her fifth hard question. Maya reaches for the approved message, and it does not cover this case, because nobody who wrote it had met Priya. So Maya impro‐ vises, Daniel hears a promise he cannot keep, and the gap opens again in front of the customer. With the session behind them, Maya does not need a message. She knows why the thesis is true, where it stops being true, and what Daniel will say about it, because they fought it out the week before. 137

Page 140 of 214. Thesis-Based Selling There is a second gain. The session moves the conflict between the two roles to a place where it is useful. By the time the customer hears a thesis, the engineer’s caveats and the seller’s ambition are both already inside it. The thesis claims exactly as much as the engineer can deliver and exactly as much as the seller needs the buyer to hear. That balance is very hard to strike live, in a meeting, under pressure. It is easy to strike in a room with only the two of you. “ A thesis has to be technically true and commercially meaningful. Who says what in the room Once the page exists, split the work. The seller states the thesis and what it means for the customer’s business. The engineer defends it with the evidence, the limits, and the conditions under which it would fail. Either one may concede a point, and when one does, the other backs it. One rule matters more than the rest. Never correct each other in front of the customer. If a thesis turns out to be wrong in the meeting, concede it together and repair the page afterward. Rewind the call with Priya. This time Maya and Daniel have agreed on a thesis, the fifth from the last chapter: what you give AI is worth more than what AI gives you. Priya asks whether it will get claims questions right. Maya says, “We think the honest answer starts with what you give it. Daniel, do you want to take that?” 138

Page 141 of 214. Agree with Your Own Team Before the Customer Daniel says, “It will be as right as the definitions and examples it is given. For the questions where you already have a known right answer, we can measure it together in the first two weeks, and I would not trust any number before that. Which claims questions have an answer you would stake a deci‐ sion on?” Same two people, same product. Priya hears one voice, and she has been handed a way to find out for herself. Walk back through what changed between the two calls. In the first, Maya answered from outcomes and Daniel from conditions, live, in front of the customer, and Priya heard two companies. In the second, they had already argued it out. Maya did not try to answer a technical question herself; she named the thesis and handed it to Daniel. Daniel did not hedge; he gave the condition and, in the same breath, a way to test it together. Nothing about the product was different. What was different was that the seller and the engineer had agreed, before the call, on what was true and on who would say it. That is the whole lesson of the chapter in one exchange. The gap between your seller and your engineer is real, and it is useful. Find it in a room where it costs nothing, turn it into a thesis you both believe, and the customer will hear the result as confidence. The phrases that carry the agreement Agreement with a customer is built in a few sentences, said at the right moments. These are the ones I have seen work, first inside the team and then in the room. Inside the team, before the deal: “Before we build the deck, let’s each write three things we believe about this market. No products.” “Is this true on real data? Under what conditions would it break?” “Suppose it is true. Why would their chief financial officer care?” “What do we not know yet, and will we both say so?” With the customer, in the first meeting: 139

Page 142 of 214. Thesis-Based Selling “Before I show you anything, here are three things we believe are true about where this is going. Which ones do you not buy?” “Is that how your leadership sees it, or how you see it?” “Who in your company would push back on this one?” “If this held, what would you stop doing?” And when one of you is about to drift: “Daniel, do you want to take that?” (hand the question to the person who owns the answer) “That is one of the things we have not seen enough of to say. Here is how we would find out together.” Each phrase does one job. The internal ones force the argument to happen early. The customer ones invite disagreement instead of waiting for it. The last two keep the voice single when the pressure is on. When agreement goes wrong Several things go wrong here often enough to plan for. The seller overreaches in the room. Maya, under pressure, says the accuracy is “very strong.” Daniel’s instinct is to correct her on the spot. Do not. Correcting each other in front of the customer is the fastest way to lose them. Let the moment pass, and repair it before the call ends: “Let me be precise about that. It depends on the definitions, and here is how we would measure it.” Then fix the page afterward, together. The customer nods at everything. A buyer who agrees with all of your theses is not agreeing. They are being polite, or they have not engaged. Ask the question that breaks the politeness: “If you had to cross one of these out, which would it be?” Most buyers will cross one out, and the reason is your next discovery note. Only the champion agrees. Your champion loves the theses, and nobody else in the buying group has heard them. That is a single point of failure. Ask the champion who would push back hardest, and ask for twenty minutes with that person. Agreement that exists only in your champion’s head will not survive the budget meeting. 140

Page 143 of 214. Agree with Your Own Team Before the Customer The account team quietly disagrees. Sometimes the thesis session ends with the seller and the engineer still not quite aligned, and both sign anyway because the meeting is tomorrow. Do not sign. Write the disagreement down as one of the unknowns and agree that you will both say so. An unknown said honestly is far less damaging than a disagreement the customer finds for you. Each of these failures has the same root: agreement that was assumed rather than tested. The cure is the same each time. Say the thesis out loud, invite the person most likely to disagree, and write down what they say. FIELD NOTE 08 The customer will find the gap between your seller and your engineer. Find it first, in a room where it costs nothing. 141

Page 144 of 214. Thesis-Based Selling YO U R TAS KS What to do this week Book forty-five minutes with the person on the other side of your deals, the engineer if you sell or the seller if you are the engineer. Call it a thesis session, and book it before the account plan and the deck, not after. Run the three steps on one live account: write separately, attack each other’s list, keep what survives and write down what you do not know. Expect it to be uncomfortable. That discomfort is the gap the customer would otherwise have found. Sign the page. Put both names at the bottom. A signed page is a commitment that you will say the same thing in the room, including “we have not seen enough to say.” Use it once. Open your next customer meeting from the page, and note which thesis drew the first objection. Bring that note to your next session; it is the best feedback the page will get. 142

Page 145 of 214. Chapter Nine Build the Sale Up from What Is True “Give me a place to stand, and I shall move the earth.” Attributed to Archimedes IF THIS IS TRUE A buyer who agrees with your theses will reach your conclusion without being pushed. THEN The work of the sale moves from persuading to agreeing on what is true, one layer at a time. 143

Page 146 of 214. Thesis-Based Selling The last chapter ended with one voice inside your team: a page of theses that the seller and the engineer have argued over and both signed. That matters, but it is still only your side of the table. A thesis does not carry a deal until the buyer has heard it, pushed on it and agreed to at least part of it. This chapter takes the theses to the buyer. It shows how to open the first meeting by agreeing on what is true, how to treat every disagreement as discovery, how to build the sale upward, layer by layer, until the product is the last thing on the table, and how to turn the agreed stack into a document the buying committee can decide and govern with, in place of the RFP checklist. This is the part of the method that replaces the old path’s demo-first meeting with something the buyer can check for themselves. Around 300 BC a Greek mathematician named Euclid wrote the Elements. It opens with a handful of statements so plain that nobody would refuse them: five postulates and five common notions. A straight line can be drawn between any two points. Things equal to the same thing are equal to each other. From those he built thirteen books and 465 propositions, each one resting on the ones before. The method mattered more than any single result. A reader did not have to trust Euclid. They could check the starting statements, then check each step, and arrive at the conclusion under their own power. For two thousand years it was the pattern for how to convince a person of something without asking for faith. A seller in a market without references needs the same thing. The buyer cannot take your word for it and has no neighbor to copy. So give them starting statements they can check and steps they can follow. This chapter finishes the first step of round two, agreeing with the buyer, and takes the second, Build. The four layers of a sale, built from the bottom A sale built this way has four layers, and it gets built from the bottom. The thesis. A claim about the world. Take my fifth: what you give AI is worth more than what AI gives you. The implication. What that means for this company. Then, because AI models keep getting better, the answers we generate this year will be generated 144

Page 147 of 214. Build the Sale Up from What Is True again, and what lasts is what we gave: our prompts, our data, our definitions, our intent. The decision. What to do first. Then we should write down our prompts, definitions and intent before we choose any tool, and keep them where every better AI model can use them. The product. What to buy. Then we need something that looks after our prompts and gives us access to the AI models. Only now does anything with a price appear. I call these four layers, built in this order, the stack. The rest of the book uses the word for exactly this: thesis, implication, decision, product. PRODUCT DECISION what to do first IMPLICATION what it means for this company BUILT FROM THE BOTTOM what to buy THESIS a claim about the world Figure 1 9 The buyer agrees to the bottom layer first. The price shows up last. Most selling runs the other way. It starts at the top with the product and works downward, looking for reasons. The demo is the purest case: forty minutes of product before anyone has agreed on what is true. The buyer knows this and discounts every reason accordingly, because each one arrived after the conclusion. Start at the bottom and the order of trust flips. The thesis costs the buyer nothing to consider, because it is about the world and would be true with any vendor. By the time a product appears, it is sitting on three layers the buyer has already accepted. 145

Page 148 of 214. Thesis-Based Selling Open by agreeing on what is true Lawyers have a useful habit, and it is worth understanding properly, because it is the pattern for how a thesis-based meeting opens. Before a trial, the two sides stipulate. A stipulation of fact is a written agreement between opposing lawyers that certain facts are not in dispute: the contract was signed on this date, by these people, for this amount. In the United States the federal rules even ask the judge, in the conference before trial, to look for exactly this: admissions and stipulations about facts and documents, so that nobody has to spend court time proving what nobody contests. Lawyers who are about to fight each other do this for good reasons. It saves days of testimony about things nobody doubts. It removes the noise, so the judge or the jury can see the real question. And it builds a record that neither side can walk back later: once a fact is stipulated, it is settled for the rest of the case. Nobody argues about the date of the contract or who signed it. The argument is saved for the questions that decide the case. A sale can open the same way. Agree on what is not in dispute first, and spend the meeting on what genuinely is. Here is what we heard, and here is what we believe This is not a book about demos. But anyone who has watched a good sales engineer open a demo or a presentation has seen the first version of this idea. Before showing anything, they put up a slide that says, in effect: here is what we heard. The three or four things the customer said in discovery, played back in their words. Then the question: did we get that right? Then the agenda: here is what we will cover today, does that work for you? Only then does the demo start. That is table stakes. It is the first thing good sales engineers are taught, and it works, because it gets the buyer to agree on the starting point before the seller builds on it. A thesis takes it one step further. Alongside what you heard, you lay out what you believe: the theses the rest of the presentation rests on. Then you ask whether the buyer agrees with those too. Here is what that opening can look like, for a meeting with the insurer from Chapter 8. 146

Page 149 of 214. Build the Sale Up from What Is True What we heard. Your analysts have a three-month backlog. Legal has approved one AI model. The last two reporting projects stalled on the defini‐ tions of a claim, a policy and a loss. Did we get that right? What we believe. The definitions will outlive any tool you buy. Better AI models will keep arriving, so anything built around one AI model will be rebuilt. In a regulated company, approval moves more slowly than the AI models do. Which of these do you not buy? What we will cover today. How to keep the definitions outside any one AI model, what the first two weeks could measure, and what we would need from your legal team. Does that work? The middle part is the one that matters. If the buyer disagrees with what you believe, the rest of the presentation will be off, however good the demo is, because every screen you show is built on a belief they do not share. So that disagreement is the most valuable thing to surface, and the earlier the better. When it comes, pause. Do not push on to the demo. Have the thesis conversation: ask what is different for them, listen, and either defend the thesis, draw a boundary on it, or concede and adjust what you were about to show. A twenty-minute detour at the start saves an hour of demo that would have landed on the wrong foundation. This is the Agree step, done in front of the customer, and it is the bottom of the four layers: you are checking the foundation before you build on it. One caution about the middle part. Chapter 5 showed that one deal holds many theses, because each role sees the same purchase from a different chair. If the room holds the chief data officer and the security officer, the “what we believe” slide needs the shared thesis both can accept, and then a line each of them recognizes as their own: for Priya, that the definitions are her asset; for the security officer, that nothing touches claims data without controls he can audit. I prepare that slide for the people actually in the room, not for the market in general. When a new role joins the next meeting, I add their line and ask the question again. Most first meetings in enterprise sales do the opposite. The seller opens with the company story, the logo slide and a demo, and the buyer spends an hour watching for the moment the pitch turns into something they can use. Nothing has been agreed, so nothing can be built on. Open a first meeting the way a lawyer opens a trial: by finding out what both sides accept. 147

Page 150 of 214. Thesis-Based Selling THE QUESTION “Before I show you anything, here are three things we believe are true about where this is going. I would like to know which ones you do not buy.” Then stop talking. This is harder than it sounds. The silence after a question like that feels long, and every seller’s instinct is to fill it with a slide. Let the buyer fill it. What they say first is usually the most honest thing you will hear all quarter. Three things can happen with each thesis. The buyer agrees. Move up a layer. The buyer disagrees, and you have an answer. That is a defense, and Chapter 10 covers it. The buyer disagrees, and they are right. Then you revise the thesis, or you learn that this buyer does not share the foundation your product is built on. The second outcome is worth having in week one. MEDDIC would call it qualification. You have saved a quarter. Disagreement is the new discovery Standard discovery asks the buyer about their pain. It works when the buyer has lived with the problem long enough to describe it: a slow report, a system that falls over every month-end, a team that cannot keep up. With AI many buyers have not lived with the problem yet. Ask them about their pain and they answer with what they read last week, or with what the last vendor told them. Lay down a thesis and the buyer has something concrete to push against. When Priya says, “That may hold for a software company, but our legal team has approved exactly one AI model and we will be on it for years,” you have learned more about her constraints from one sentence than an hour of open questions would produce. The objection shows where her world differs from the general case, and that difference is where the deal gets decided. Listen for three things in a disagreement. What the buyer cannot change, like a regulator or a contract. What the buyer is afraid of, like a wrong 148

Page 151 of 214. Build the Sale Up from What Is True number reaching the board. And who else is in the sentence, like the legal team Priya mentioned. Each one is a line in your discovery notes that no open question would have produced. Why being corrected beats being asked The methods most sellers were trained on find the pain first. Solution Selling builds a chain from one person’s pain to the next; Gap Selling describes the buyer’s current state before the future one, as Chapter 4 explains. Both lean on the open question, and as the last section showed, the open question comes back nearly empty when the buyer has not yet run AI at scale. Why does a thesis do so much better? There is a better way in, and it rests on something everyone has seen online. Ward Cunningham, who invented the wiki, has a law named after him, coined by a former Intel executive, Steven McGeady, in 2010: the best way to get the right answer on the internet is not to ask a question; it is to post the wrong answer. Ask a forum an open ques‐ tion and you may wait for days. Post something that is clearly wrong and someone will correct you within minutes. I have used this for years to get feedback, in meetings and outside them. Ask a room “what are your thoughts?” and it goes quiet. Say something specific that might be wrong and people tell you, instantly and in detail. Post a vague question on LinkedIn and it sits there. Post a confident claim that does not quite hold and the comments fill up with people setting you straight. People will not always volunteer what they know. They will almost always correct a mistake. That instinct is why a thesis works so well in discovery. A thesis is not a vague point of view. As Chapter 5 defines it, it is a specific, grounded claim about the world, low enough to the ground that a person can see exactly where it is wrong for them. Lay one down and the buyer has something concrete to correct. Priya’s line about her one approved AI model was exactly that: a correction, and the correction was the discovery. One line matters here. State your honest best thesis, the one you actually believe, not a wrong one planted to bait a reaction. Cunningham’s law is a mechanism, not a trick. Buyers can tell the difference, and a seller who plays games with the truth loses the trust the whole method depends on. Do not be afraid of being wrong. You might think it is bad to walk in with a thesis that turns out to be wrong and have to defend it. You will be 149

Page 152 of 214. Thesis-Based Selling surprised how much it uncovers. Even if your thesis is completely wrong, the correction is discovery: you take it back, rebuild the thesis, and come back with one that has been tested against a real buyer. Every correction is ammu‐ nition. And the good ones compound. A thesis that is grounded in what is true does not stay true for only one company. If it holds at one insurer, it will tend to hold at the next, or for the same role in a different industry. So the theses you test and repair become portable, an asset you carry from deal to deal and, as Chapter 5 said of the best sellers, from company to company. That is why it is a win either way: right, and you have a foundation; wrong, and you have a better thesis. I have watched sellers who believe their job is only to ask: what do you want to do, what is your opinion, what is your pain. That works when the customer knows the answer. With something this new, they often do not, and that is exactly the problem you are there to help them solve. YOU STATE THE THESIS AGREES with the thesis DISAGREES you have an answer DISAGREES they are right Move up a layer Defend it: Chapter 10 Revise, or qualify out Figure 2 0 Three things can happen with each thesis. Each tells you where to go next. Notice that every path in the figure leads somewhere useful. None of the three outcomes is a failure, and that is the difference from a demo, where the only good outcome is that the buyer likes what they see. The only meeting that goes nowhere is the one where the buyer was never asked. “ A thesis gives them something to push against. 150

Page 153 of 214. Build the Sale Up from What Is True Let the buyer say the next line Each step up the stack should be small enough that the buyer says it before you do. THE QUESTION “When a better AI model arrives, what in your plan would have to be rebuilt?” Priya thinks. “Most of it. We wrote the prompts around one vendor.” “And what would carry over?” “The data. The definitions of a claim, a policy, a loss.” She has stated the implication and half the decision. She will defend those in her own budget meeting, because they are hers. Look at what the seller did in that exchange. Two questions, no claims. The thesis was already agreed, so each question only asked Priya to look one step up from it, and the step was small enough that she could see the answer herself. Had the seller said, “So you should store your definitions outside any one AI model,” Priya would have heard a pitch and started weighing it. Because she said it, there is nothing to weigh. People hold a conclusion more firmly when they reached it themselves, and a conclusion the buyer reached is one she will repeat in rooms you never enter. The seller’s job here is to keep the steps small. If the buyer cannot say the next line, go back down a layer and ask a smaller question. The same order shapes the proof of concept. In the old path, a trial showed that the product did what the demo promised. Here the trial tests the thesis. Daniel’s offer to Priya in Chapter 8 was exactly that: take the ques‐ tions that already have a known right answer, run them for two weeks, and count. The measure comes from the thesis, not from the feature list, so whatever the result, both sides learn something they can build on. 151

Page 154 of 214. Thesis-Based Selling The stack has to travel without you Chapter 5 put a number on this: buyers spend only about 17 percent of their buying time with suppliers. The rest happens in rooms you are not in, among the buying group Chapter 3 described, each member reading some‐ thing different. A demo cannot go into those rooms. A deck goes in and gets skimmed. A stack of three sentences goes in whole, because your champion can say it from memory: this is true, so this follows, so we should do this first. When someone in that room pushes back, the champion can lay out each step and ask which one is wrong. That is a far stronger position than “the vendor said so.” In a mature market the champion would carry a reference into those rooms: a company like ours did this, and it worked. With no reference to carry, the stack is what the champion carries instead, and it holds up for the same reason the reference did, because the people in the room can check it themselves. FIELD NOTE 09 A demo stays in the meeting. An argument the buyer can repeat goes everywhere the decision goes. Replace the RFP checklist with a thesis memo A stack in the champion’s head travels well. A stack on paper, in the buying committee’s own file, travels further, and it does a job nobody else in the deal is doing. Chapter 2 named the procurement time trap: the capabilities move faster than the questions, and faster than the process that asks them. Chapter 5 showed that a company committing to foundational AI is in the position of a lead investor, and needs an investment memo rather than a scorecard. This section puts the two together, because the committee still has to decide something, compare vendors and answer for the result, and if the RFP cannot carry that, something has to. What carries it is the thesis, written as a buying document. On a seller’s slide, a thesis is part of the pitch. In the committee’s file it becomes some‐ thing else: a record of what bet the company is placing, why, and under what 152

Page 155 of 214. Build the Sale Up from What Is True conditions the bet succeeds or fails. I call that document a thesis memo. Here is one being written. After the first meeting in this chapter, Priya’s head of procurement does what procurement is paid to do. He asks for an RFP, and he already has a draft: two hundred questions copied from the last analytics purchase, most of them about features. Priya knows the draft will score the vendors on the wrong things, but she also knows procurement needs something it can defend. So she asks Maya and Daniel to help her write a two-page memo instead, with the committee’s names on it, not the vendor’s. Over a week, and two more meetings, it comes out like this. The bet. We will keep our definitions of a claim, a policy and a loss outside any one AI model, and buy only tools that can read them, so that every better AI model can use them without a rebuild. Why we believe it. Three theses, each argued over in this room. Our defi‐ nitions will outlive any tool we buy. Better AI models will keep arriving, so anything built around one AI model will be rebuilt. Approval in a regulated company moves more slowly than the AI models do. It succeeds if. In the first two weeks, on the claims questions that already have a known right answer, the answers reach a level the claims team sets before the trial starts. By the end of the first quarter, a second approved AI model can run on the same definitions without rewriting them. It fails if. The definitions have to be rewritten for each AI model, or the claims team cannot tell a right answer from a wrong one, or legal begins approving new AI models in weeks rather than months, which would undo the third thesis and change the plan. What we keep if it fails. The definitions stay ours, in an open format. We lose two weeks, not three years. When we look. Week two, the end of the first quarter, and ninety days before renewal, each time against this page. What stays a checklist. Security controls, where the data is kept, audit and contract terms. These are stable, so they are scored the usual way. The head of procurement reads it and asks the right question: how does he compare vendors against a memo? The answer is in the memo. Every vendor gets the same page and three requests: say which of the theses you share and where you disagree; say what you would measure in the first two weeks; and say how the company leaves if the bet fails. Three vendors answer. 153

Page 156 of 214. Thesis-Based Selling One answers the first two well and skips the third. That tells the committee more about lock-in than any line of the old draft would have. Look at what was done there, because it is the whole method in a single document. The stack from the start of this chapter, thesis, implication, deci‐ sion, was written down by the buyer, in the buyer’s words. The falsifiers from Chapter 5 became the “it fails if” lines. The checkpoints became dates the committee owns. And the old RFP did not vanish; it shrank to the part of the decision that really can be specified in advance. This matters for three reasons. First, the committee knows what it is buying: not a product, but a bet with stated conditions, which is what Chapter 5 said foundational AI is. Second, the memo governs the purchase after the signature. At week two and at the end of the quarter, nobody argues about what was promised, because the promise is on the page, and nobody can quietly rewrite it. Third, it is fair, and buyers can feel that. The theses are about the world, so any vendor can answer them, and Maya’s team is glad they can, because their own answers have already been tested in the room. There is one honest warning. You cannot write this memo for the buyer and hand it over, or it becomes your pitch on their letterhead. Offer the structure, argue over the theses, draft lines when asked, and let the champion and the committee own the words. Then use it. The next time a deal you are working is heading for an RFP, ask the champion whether the committee would rather decide on a short memo of what it is betting on, and offer to help write it. The workbook’s Thesis Sheet is a good first page. 154

Page 157 of 214. Build the Sale Up from What Is True YO U R TAS KS What to do this week Build one stack. Take the thesis you kept from Chapter 5. Under it write the implication, the decision, and only then the product. Check the steps. Read each “then” aloud. If a skeptical listener could not have said it themselves, the step is too big. Split it. Open one meeting by stipulating. Three theses, then the question: which of these do you not buy? Write down the first disagreement. Word for word. It is your best discovery note of the month. Offer a thesis memo. On the next deal heading for an RFP, ask your champion whether the committee would rather decide on a short memo of what it is betting on, and offer to help write it. 155

Page 158 of 214.

Page 159 of 214. Chapter Ten Defend the Thesis, and Let the Buyer Attack It “He who knows only his own side of the case knows little of that.” John Stuart Mill IF THIS IS TRUE A thesis is worth what it has survived. THEN Pushback is the test you want, and conceding a point in the open makes the next defense stronger. In the last chapter we built the sale up from what is true. You opened with your theses, the buyer pushed on some of them, and together you stacked 157

Page 160 of 214. Thesis-Based Selling the implication, the decision and finally the product on the ones that held. If we stopped there, you would already be selling differently from most of the people you compete with. But a stack is only as strong as its bottom layer, and every serious buyer will test that layer, sometimes in the meeting and often in rooms you are not in. Some of those tests will land. This chapter is about the last step of round two: taking the pushback, telling a challenge you can answer from one where the buyer is right, conceding in the open, and keeping a record that becomes the reference the market could not give you. Charlie Munger, Warren Buffett’s longtime partner at Berkshire Hath‐ away, gave a rule for holding opinions in a commencement speech in 2007. He was not entitled to an opinion, he said, until he could state the argu‐ ments against his position better than the people on the other side. Hold a thesis to that standard. Before it goes in front of a customer, you should be able to make the case against it better than the customer will. In my experience this is where the work really is. Writing a thesis takes an after‐ noon. Knowing its weakest point well enough to say it first takes practice, and it is what makes a buyer trust the rest of what you say. Defend is the last step, and the one sellers are least prepared for. In the room it usually falls to the sales engineer, because the challenge usually arrives as a technical question. Sales engineers, like most sellers, are trained to handle objections, and handling an objection usually means making it go away. Remember Cunningham’s law from the last chapter: people will correct a claim far more readily than they will answer a question. Defending a thesis is what you do with those corrections, and it is a different activity from objection handling, closer to the doctoral exam from Chapter 5. The other person is testing the claim, and you want the test to be a real one, because a claim that passed a real test is the only thing you can offer in place of a refer‐ ence. Three kinds of pushback Almost every challenge to a thesis is one of three. Take my first thesis, every employee will have an AI, and listen to three buyers push on it. “That is not true.” A challenge on the facts. “Half our people work on a factory floor and never open a laptop.” Answer with evidence, and say how 158

Page 161 of 214. Defend the Thesis, and Let the Buyer Attack It good the evidence is. Then give the falsifier: “Here is what would change my mind. Have you seen it?” Sometimes they have, and you have learned some‐ thing. “True in general, not for us.” A challenge on scope. “That may be right for a bank. We run hospitals.” This one is a gift. Ask what is different about them and listen, because they are about to tell you their constraints. Often the thesis survives with a boundary drawn on it: true for people who work at a desk, not yet for those on a hospital ward or a factory floor. “True, not yet.” A challenge on timing. “Maybe in five years.” Leave the date alone. Ask what they would need to see for “yet” to arrive, and agree on that signal together. You have turned a stall into a tripwire both sides are watching. YOUR THESIS FACT SCOPE TIMING That is not true. Not for us. Not yet. Give the evidence, then your falsifier Ask what is different, then draw the boundary Leave the date. Agree on the signal to watch Figure 2 1 Sort the challenge first. Each kind has its own answer. The figure is a sorting rule. Before you answer a challenge, decide which of the three it is, because each kind has its own answer and the wrong answer makes things worse. Meet a challenge on scope with evidence and you have argued past the buyer’s real point. Meet a challenge on timing by defending the date and you have started a fight nobody can win. Sort first, then answer the challenge that was actually made. Say what would change your mind In 1974 the physicist Richard Feynman gave the commencement address at Caltech, and he used it to warn the graduating scientists about a habit he called cargo cult science: research that has the form of science but fools its 159

Page 162 of 214. Thesis-Based Selling own author. The first principle, he told them, is that you must not fool yourself, and you are the easiest person to fool. A seller has the same problem. After a quarter of telling a thesis to buyers, it starts to feel like a fact. A written falsifier is how a seller keeps Feynman’s rule: it names, in advance, the evidence that would prove you wrong, so you cannot quietly move the goalposts later. Sellers almost never tell a buyer what would prove them wrong. It feels like handing over a weapon. In practice it does the opposite. A person who can name the evidence that would sink their claim sounds like someone who has thought about it, and a buyer who has sat through a month of vendors hears the difference at once. Here is what it sounds like. A seller lays down the first thesis, that every employee will have an AI, and adds without being asked: “I would stop believing this if companies ran the trials, found the assistants did not pay for their seats, and took them away from most of their staff. I have not seen that yet. Have you?” The buyer now knows three things: what the seller believes, what would change their mind, and that the seller is watching for it. That is more than any reference call could tell them. Write the falsifier next to each thesis, and say it in the meeting without being asked. Lose an argument well, and win the buyer’s trust Sometimes the buyer is right. The forecaster Paul Saffo has a phrase for how to hold a view about an uncertain future: strong opinions, weakly held. Commit to the claim firmly enough to act on it, and drop it the moment the evidence turns. Both halves matter to a seller. Without the strong opinion you have nothing to lay on the table. Without the weak hold you have a thesis you will defend past the point of reason, and the buyer will watch you do it. So when a buyer lands a point, say so in the meeting, in plain words. “You are right, that thesis does not hold for a company in your position.” Write the revision down where they can see it. At the next meeting, tell them what you changed because of them. Conceding costs less than sellers fear. The buyer has watched you change your mind for a good reason, which is the best evidence they will get that 160

Page 163 of 214. Defend the Thesis, and Let the Buyer Attack It your remaining claims are held for good reasons too. It also changes the buyer’s role: they are no longer the audience for your argument but one of its authors, and people defend what they helped build. Investors know this. Union Square Ventures, the venture firm whose one-line thesis we met in Chapter 5, published each new version of its thesis as it changed, and nobody thought less of the firm for it. A seller who has never revised a thesis is reciting one. Keep a ledger of every challenge Keep a record, one line per challenge: the thesis, the date, who challenged it, their argument, the kind of challenge, and what happened. There are four outcomes. It held. It held with a boundary. It was revised. It was retired. Over time the ledger holds theses that twenty careful buyers have tried to break, with a note on each attempt. When the next buyer asks who else is doing this, you still cannot point at a production customer. You can say that the claim in front of them has been attacked by twenty companies like theirs, and show what survived and what changed. THESIS DATE WHO THEIR ARGUMENT WHAT HAPPENED No. 5 March the machine most output is used once held, with a boundary No. 1 April a hospital group not true for hospitals revised No. 1 May a manufacturer maybe in five years held held held with a boundary revised FOUR OUTCOMES retired Figure 2 2 One line for every challenge, with its kind and its outcome. Read the three sample lines. In March the machine attacked the fifth thesis, and it held with a boundary. In April a hospital group challenged the first thesis on scope, and it was revised. In May a manufacturer said “maybe 161

Page 164 of 214. Thesis-Based Selling in five years”, a challenge on timing, and the thesis held. Each line is small. Twenty of them, from twenty different buyers, are something no logo slide can offer: a record of the claim being tested by people like the next buyer, and of what survived. FIELD NOTE 10 Say the falsifier before the buyer asks for it. It is the cheapest credibility in selling. Never bend the thesis to save a deal There is one rule in this chapter I would ask you to hold onto above the others. Defend the thesis and the product separately, and never bend the thesis to rescue a deal. “ Bend the thesis to save a deal, and you lose both. The temptation is real, and it arrives at the worst moment. The quarter is ending. The deal is slipping. The buyer pushes back on one of your theses, or a competitor’s product is cheaper or already approved. And the seller starts, quietly, to reshape the thesis. It does not really apply here. Or it applies, but only in a way that happens to favor our product. Or it becomes, word by word, whatever the buyer said in the last meeting. It fails, and it fails for a reason worth understanding. The thesis is loadbearing. Everything in the deal stands on it: the implication, the decision, the case for the product, the criteria the buyer will judge by. Bend the founda‐ tion and everything above it shifts. The decision no longer follows. The criteria no longer fit. The product case floats. Buyers sense it fast. A thesis that suddenly matches whatever the buyer said last does not sound like a belief; it sounds like a sales tactic, because it is one. It usually loses its falsifier on the way, because a thesis bent to fit the buyer can no longer say what would prove it wrong. The champion, who has been carrying your argument into rooms you are not in, can no longer retell a story that keeps changing. And once the buyer has watched you bend one 162

Page 165 of 214. Defend the Thesis, and Let the Buyer Attack It claim, they discount every other claim you make. You lose the thesis and the deal. There is a difference between bending a thesis and revising one, and it is the difference between weakness and strength. Here are the two side by side. Revising. A buyer at a hospital group challenges your first thesis: every employee will have an AI. Their nurses spend the shift on the ward, not at a screen. You agree, in the meeting, that the thesis does not hold there yet. You redraw it: every employee who works at a desk will have an AI. You write the change in the ledger, with the date and the reason. At the next meeting you tell them what you changed because of them. The thesis is narrower, better tested and still yours. Bending. A different buyer says their budget is frozen and they want to wait a year. Instead of asking what signal would end the wait, the seller decides that maybe the thesis is really about next year, and maybe it is mostly about cost savings, which is what the buyer seems to care about, and maybe the product can be sold as a cost tool after all. Nothing in the evidence changed. Only the quarter did. The ledger shows nothing, because there is nothing honest to record. So when a thesis is under pressure, you have four honest moves. Defend it, if it is true, with the evidence and the falsifier. Bound it, narrowing the scope honestly, as the hospital buyer helped you do. Concede it in the open, if the buyer is right, and repair it where they can see. Or walk away, if the buyer’s theses and yours do not overlap and the conversation about the thesis cannot move them. Each of those keeps the foundation sound, what‐ ever happens to this deal. If a buyer accepts your thesis and still prefers a competitor’s product, the thesis is intact. You lost at the top layer of the stack, and the foundation you built is still in that account for next year. Lose that argument honestly, and the buyer remembers the seller who told the truth about the world even when it did not help the deal. In a market without references, that memory is worth more than most wins. All of this is why the thesis deserves the same effort you give the product. Sellers spend weeks learning a product: its features, its limits, its roadmap. A thesis that the whole deal rests on deserves at least as much. Research it. Test it against the machine, your counterpart and a friendly customer. Write down what would prove it wrong. Do not make it up the night before a meeting, because whatever you make up, you will be tempted to bend. 163

Page 166 of 214. Thesis-Based Selling YO U R TAS KS What to do this week Argue the other side. For each of your theses, write the best case against it, the one a sharp buyer would make. If you cannot, find someone who can, because a thesis you cannot attack is one you do not yet understand. Sort your last five objections into fact, scope and timing. Note which kind you hear most. If it is mostly timing, your theses may be right and your signals missing; if it is mostly scope, your theses may be too broad. Say a falsifier out loud in one customer meeting, before anyone asks for it. Watch how the buyer’s posture changes when they hear you name what would prove you wrong. Start the ledger. A spreadsheet with six columns will do: the thesis, the date, who challenged it, their argument, the kind of challenge, and what happened. 164

Page 167 of 214. Chapter Eleven Use AI to Attack Your Thesis Before the Buyer Does “Weak human + machine + better process was superior to a strong computer alone.” Garry Kasparov, 2 0 1 0 IF THIS IS TRUE A language model is the cheapest hostile committee you will ever have. THEN Every thesis should lose an argument to a machine before it meets a customer. 165

Page 168 of 214. Thesis-Based Selling Defending a thesis in front of a buyer is the real test, and it is expensive. Every hole the buyer finds costs you credibility in a deal that matters. The last chapter showed how to take those hits well. This one shows how to take most of them earlier, in private, from an opponent that never tires and never spares your feelings: a language model. The hard part of defending a thesis used to be finding someone to attack it. You needed a skeptical colleague with a free hour, and most colleagues are too busy or too polite. That constraint is gone, which means you can rehearse the last step before a customer makes you do it for real. A language model will play a skeptical chief financial officer, a security architect, or your competitor’s best engineer, at eleven at night, as many times as you ask. It does not get tired and it does not spare your feelings. AI can attack your thesis. It cannot give you one. Before the prompts, one line needs drawing, and I want to draw it as clearly as I can. Use AI to challenge your thesis, to stress-test it, to find the counter‐ arguments and to gather the evidence. Do not use it to originate the thesis. The reason is in how these AI models work. A language model is trained on what has already been written, and when you ask it a question it tends toward the most likely answer: the consensus, the sensible middle. Ask it for a thesis about AI in insurance and you will get a reasonable sentence. It will be the sentence every other seller gets when they ask the same question. Research on writing with AI has found the same pull: in one study pub‐ lished in Science Advances in 2024, stories written with an AI model’s help were rated better individually, and were more similar to each other than sto‐ ries written without it. Your competitors have the same AI models you do, and so does your buyer. Forrester’s 2024 survey of business buyers found 89 percent of them using generative AI as a source of information during a pur‐ chase. If your thesis came from an AI model, assume your competitor’s did too, and assume the buyer has already read it. Average in, average out. An AI model also has nothing at stake. It will argue either side with the same fluency, which makes it an excellent opponent and a poor author. A thesis is a position a person takes and answers for, and the answering cannot 166

Page 169 of 214. Use AI to Attack Your Thesis Before the Buyer Does be handed off. And an AI model invents: any fact, number or source it offers in an argument has to be checked before you repeat it. What stays with you is the part that matters most. Conviction. A point of view grounded in what you have seen: in your deals, your team’s experi‐ ence, your product’s knowledge of this kind of buyer. That is the work that grows in value as AI does more of everything else, and it is the last thing to delegate. I know how this sounds in a book that spends so many pages on how capable AI is becoming. The claim will be challenged, and maybe one day an AI model will do more than inform and point. Even then, the thesis has to be yours. A thesis is a belief you are willing to defend, and you cannot defend a belief you did not form. It is also, as the Introduction said, where the energy in a sale comes from: a buyer can feel the difference between a seller who believes something and a seller who is reading an AI model’s answer. There is a deeper reason, and it is one of my own five. The fifth thesis in Chapter 7 says: what you give AI is worth more than what AI gives you. The value is in what you bring to the AI model: your context, your judgment, what you know about this buyer, and above all your thesis. The AI model’s output is only as good as that input, and it will be replaced by a better AI model’s output next year. What you gave it is the part that lasts. So the thesis is the most valuable thing you will ever hand an AI model, and it has to come from you. Picture two sellers preparing for the same account. The first opens a chat window and types, “What should my thesis be for this insurer?” The answer is fluent, sensible and the same one the competitor received an hour earlier. The second takes a piece of paper and writes three theses about this insurer, from what she has heard in discovery and what she has seen at companies like it. Only then does she open the AI model, paste the three in, and ask it to attack them. The first seller has an average thesis and no reason to believe it. The second has her own, tested, and she knows exactly where it is weak. So here is the practice, as plainly as I can say it. Take a piece of paper. Write your thesis before you open any AI tool, however tempting it is to start there. Then bring it to the machine to sharpen. The five dated bets you wrote before you started reading are the same kind of thing: your own calls about the future, on paper, in your own words. Let the machine attack those, and keep the paper. 167

Page 170 of 214. Thesis-Based Selling A machine cannot give you the thesis. Four prompts to attack your thesis These four cover most of the work. Use them in order. 1 PROMPT 1 OF 4 T H E AT TA C K Here is my thesis: [one sentence]. Give me the strongest argument against it. Do not soften it and do not balance it. W H AT I T D O E S It finds the hole you have not seen: the strongest case against the thesis, the kind a skeptical buyer would make in the second meeting. 2 PROMPT 2 OF 4 T H E FA L S I F I E R What evidence, if it appeared in the next twelve months, would prove this thesis false? W H AT I T D O E S It writes your falsifier for you to check: the evidence that would prove the thesis wrong, which you then say out loud before a buyer asks. 3 PROMPT 3 OF 4 THE REHEARSAL You are the chief financial officer of a mid-size insurer. Three vendors have made this claim to you this month. What do you ask the fourth? W H AT I T D O E S It rehearses the meeting. The AI model plays the role you name, and you practice answering challenges on fact, scope and timing before they are real. 168

Page 171 of 214. Use AI to Attack Your Thesis Before the Buyer Does 4 PROMPT 4 OF 4 THE TRADE Rewrite my thesis so it claims less and is harder to attack. Then tell me what I gave up. W H AT I T D O E S Shows you the trade between a bold thesis and a safe one, so that you choose where to stand on purpose. Use the four in order, and read the output with care. The first prompt gives you the attack; read it for the one point you cannot answer, and ignore the rest. The second gives you a falsifier; rewrite it in your own words until you would be comfortable saying it to a chief financial officer. The third is a rehearsal; run it until the challenges stop surprising you. The fourth asks the AI model to make the thesis claim less; take its version only if it is still worth defending. None of the four writes your thesis. They make the one you already hold harder to break. A worked example: the fifth thesis Take the fifth thesis from Chapter 7: what you give AI is worth more than what AI gives you. Asked for the strongest counterargument, an AI model will usually say something like this. Most of what people ask AI for is used once. A summary of today’s contract has its value today, and nobody will produce it again with a better AI model next year. For one-time work the output is the whole point, and the request behind it is a throwaway line. That is a fair hit. The thesis survives it, with a boundary. The revised version reads: for anything a company will build more than once, what you give AI is worth more than what AI gives you. It claims less, and it is harder to attack. It also hands you a discovery question for the next meeting: which of your work gets rebuilt every time the tools improve? 169

Page 172 of 214. Thesis-Based Selling THE THESIS what you give AI is worth more than what AI gives you THE ATTACK most output is used once before the attack WITH A BOUNDARY for anything a company will build more than once claims less, harder to attack Figure 2 3 The machine lands a fair hit. The thesis survives with a boundary. The figure shows one round of sharpening. The thesis went in claiming that what you give AI is worth more than what AI gives you, everywhere. The machine found the case where that is false: work used once. The thesis came out with a boundary, true for anything a company will build more than once. It claims less and it is harder to attack, and the boundary itself became a discovery question for the next meeting. That is what a good attack does to a thesis, whether it comes from a machine or a buyer. The routine: four attackers, in order Here is the routine I use, and I would ask you to make it a habit rather than an event. A thesis gets stronger by being attacked, and attacks have a cost. Some cost ten minutes. Some cost a favor. The most expensive cost a deal. So I put my theses in front of four attackers, in order from cheapest to dearest: the machine, my counterpart, a friendly customer, and the market. Each one finds holes the one before could not see, and by the time a thesis meets a buyer who matters, the cheap mistakes are gone. The machine. It starts on a Sunday evening, or between meetings on a Tuesday. I take the thesis I wrote on paper, paste it into an AI model, and run the four prompts above. It takes ten minutes and costs nothing. What the machine is good at is the arguments anyone could find: the standard objections, the counterexamples already in print, the places where the sentence claims more than it can carry. A typical hit sounds like the one on the fifth thesis earlier in this chapter: most of what people ask AI for is used once. I do not argue with it. I ask whether it is fair, and if it is, I narrow the thesis and write down why. My counterpart. Next comes the person who sells alongside me: the sales engineer if I am the account executive, the account executive if I am the engineer, the partner from the thesis session in Chapter 8. This costs a 170

Page 173 of 214. Use AI to Attack Your Thesis Before the Buyer Does colleague most of an hour, so I bring the thesis already sharpened by the machine. What my counterpart can see that the machine cannot is our own deals. The engineer asks whether the thesis is true on real data, at the customers we actually have. The seller asks whether a chief financial officer would care. A typical hit: “That holds for our cloud-native customers, but half our pipeline still runs on their own hardware.” We change the wording together, and we both sign the new version. A friendly customer. Then I take it to someone who will tell me the truth because they owe me nothing: a customer I have worked with for years, a former buyer who has moved companies, a peer in an industry I sell into. I do not make it a sales call, and I say so up front. “I am not selling you anything today. I am testing an idea, and you are one of the few people who will tell me if it is wrong.” Over coffee, or on a call with no slides, I say the thesis and ask where it breaks for a company like theirs. A friendly customer sees what no colleague can: the politics inside a buyer, the budget cycle, the regulator, the reason a sensible plan dies in a steering committee. A typical hit: “Our risk team would never sign off on that in year one.” That does not kill the thesis. It tells me the boundary and the timing signal to watch. The market. Only then does the thesis meet buyers who matter, and from then on every meeting is an attack. I record each one in the ledger from Chapter 10, the same day. In practice a line reads something like this: the first thesis, the fourteenth of May, the head of operations at a logistics company, “our drivers will never use an assistant in the cab,” a challenge on scope, held with a boundary: true for staff at a desk, not yet for staff on the road. One line takes two minutes. Twenty of them are a record of the thesis tested by the kind of people the next buyer will be. Look at what the routine does. Each attacker costs more than the last, and each sees something the last could not: the machine sees what is written, my counterpart sees our deals, the friendly customer sees inside a buyer, the market sees everything. By the time a thesis reaches a buyer who matters, it has been attacked three times by people with nothing to lose. That is the defense from Chapter 10 rehearsed in advance. And it is the best protection against the temptation to bend a thesis in front of a buyer, because a thesis that has already survived three honest attacks is one you have earned the right to hold. 171

Page 174 of 214. Thesis-Based Selling “ A thesis that reaches step four has already lost three arguments and been repaired three times. FIELD NOTE 11 Let the machine attack your thesis. Never let it write one for you. YO U R TAS KS What to do this week Run the four prompts on your best thesis, the one you wrote on paper before opening the tool. Run them in order, and keep the AI model’s answers. Check every fact the AI model gave you before you use it anywhere. AI models invent sources and numbers with complete confidence, and one made-up citation in front of a buyer costs more than the whole exercise gains. Rewrite the thesis once, using the strongest hit. Narrow it, bound it or sharpen it, but keep it yours. Log it. The machine’s attack is the first line in your ledger. Record the thesis, the date, “the machine,” the argument, the kind of challenge and what you changed. 172

Page 175 of 214. Closing Post Your Thesis Where Everyone Can Argue with It We have come a long way together. We started with the path enterprise selling was built on and why it is breaking. We looked at why people buy the thesis before the product, why the reference customer has gone missing, and what the methods you already know quietly assume. Then we did the work: what a thesis is and how to test one, how to find the theses under a decision, a set of my own to argue with, how to agree on them with your team and your buyer, how to build a sale on them, how to defend them, and how to let a machine attack them first. Before you put the book down, I want to bring it home, and leave you with something you can act on this week. Start where the book started, with the one question every buyer asks. THE QUESTION Who else is doing this? 173

Page 176 of 214. Thesis-Based Selling For most of the history of selling there was an answer, and it was a name. For now, in AI, there is no name to give. That will not last. Markets mature. In a few years there will be plenty of companies with long records, and pointing will work again. The references of that later market are being created now, in deals closed without any. They will belong to the sellers who could explain, step by step, why a buyer should move before there was anyone to follow, and who turned out to be right. So the answer for now is the one from the Introduction. THE ANSWER Nobody has done this long enough to prove it. Here is what we believe is true. Here is what would prove us wrong. Here is what follows for you if we are right. Tell me where you disagree. That answer does what a reference used to do. It gives the buyer something to rely on, and it is honest about what nobody knows yet. Everything in this book serves that paragraph. You write the theses down so they can be tested. You agree on them with the person who sells beside you so the customer hears one voice. You build the sale upward from them so the buyer can check every step. You defend them, concede when you should, and keep the record. I am aware of something as I write this. Thesis-Based Selling is itself a thesis: a claim that on a large decision people buy the thesis before the product, and that a seller who comes with one, honestly held and openly defended, will do better than a seller who comes with a feature list. I hold that belief. I have also told you what would prove it wrong, and I mean it. Test it the way this book asks you to test everything else. I wish the moment we are in had a name. If it needs one, call it selling before the proof: selling when the references do not exist yet, and the only thing you can bring is what you believe and why. It is uncomfortable, and it is also the beauty of this profession. When there is nothing to point at, the seller matters more, not less. 174

Page 177 of 214. Post Your Thesis Where Everyone Can Argue with It And the moment is bigger than selling. The rule books we learned from no longer hold. The paradigm has shifted so far that I am writing about selling to the enterprise while assuming that enterprises, as we know them, will still look much the same a year from now, and I am not sure they will. Some people go further and talk about where all of this ends for humanity. I will not go there. But this is a profound time to be doing this job, and I do not take it for granted that the profession of enterprise selling will survive in the shape we know. I hope it does. If it does, this is how I believe it should be done. Products will change every quarter. The AI model you sell on this year will be gone in two. The company you sell for may not exist in its current form in five. Your theses are the part you carry from deal to deal and from company to company, and they improve each time someone tries to break them. Your customers are the ones who have to turn. Nobody steers a company through a change this large by instructing each person. Like the starlings from the Introduction, their people turn together when they hold the same few things to be true. A seller who brings those few things, plainly said, is doing the work the customer most needs done. The product will soon explain itself. Nobody will need you to read them the release notes. They will need you to say what you believe about where their world is going, and to help them decide what they believe. So do not treat this as a book you have finished. I wrote it to be used. A thesis you never take into a room is worth nothing. Take one into your next meeting. Lay it down. Ask which parts the buyer does not buy. Write down what they say. And notice how far beyond selling this goes. You already live by theses. When you buy a car, you are acting on a belief about what matters in a car. When you choose where to live, you are acting on one about the life you want. When you accept a job offer, you are betting on a thesis about where that company and your career are going. Once you see it, you will see theses everywhere, and you will make better decisions for having said yours out loud. Take this method and make it yours. Extend it. Share it with someone you sell beside, and try it together. Your theses will be challenged, and you will learn some hard lessons. Some of your theses will need a boundary; some will need to be retired. That is the method working. If you keep digging 175

Page 178 of 214. Thesis-Based Selling until you reach the ground truths, the beliefs about how the world works, you will put your customers in a better position to decide, and you will build a career on conviction instead of on whichever product you happen to be selling. Sell what you believe. Make your bets, write them down with the date, defend them, and change them when the evidence says so. Over time, the bets that survive become your theses. Everything around us was built this way. Every company, every product you use, every change that once looked impossible started as a belief someone held before there was proof, and then worked to make real. That is the seller’s job too. Not to wait for the proof, but to hold a conviction worth holding, and help a customer make it come true. Luther posted his on a door, where anyone could read them and argue. Do the same. Write yours down this week. Put your name under them. Post them where the people you sell with, and the people you sell to, can read them and argue. Fru 176

Page 179 of 214. Field Guide The Framework at Work The chapters made the argument. This part is for using it. It follows one deal through all six steps, shows where each step falls in a deal, gives each step a card you can work from, and adds a scorecard for deal reviews, a page for managers and the lines worth saying out loud. The workbook after it has the pages to write on. One deal, six steps The deal below is a composite. Priya, Maya and Daniel appeared in Chapters 8 and 9; here is their whole deal, start to finish. The scene is invented, built from the kind of deal anyone who has sold AI to a regulated company will recognize. List. Maya and Daniel start before the first big meeting, on their own. At the top of a page they write the decision in Priya’s words, without a product in it: let our claims and finance teams ask questions of our claims data in plain 177

Page 180 of 214. Thesis-Based Selling language, this year. Under it they list every reason they have heard for doing it, and every reason they have heard for waiting. The reasons to act come quickly. The board has asked what the company is doing with AI. The analysts have a backlog of three months. Loss reports reach the executives weeks late. Most of Priya’s staff already use AI at home and wonder why they cannot use it at work. The reasons to wait take longer, and they matter more. Legal has approved exactly one AI model. The last two business intelligence projects stalled because nobody could agree what a claim, a policy and a loss meant. And the fear Priya mentioned once and never repeated: a wrong number reaching a regulator. Ask why. Under each reason they ask why it matters, and keep asking. The board pressure, the backlog and the late reports all come down to one belief: the company’s people need answers from its data faster than the current process can give them. The stalled projects and the fear of a wrong number come down to another: the definitions of a claim, a policy and a loss are the hard part, and they will outlast any tool the company buys. The single approved AI model comes down to a third: in a regulated company, approval moves more slowly than the AI models do. Fifteen reasons have become three beliefs. Write it down. Each belief becomes one sentence, with what would prove it wrong. Maya and Daniel also bring one of their own, the bounded version of the fifth thesis from Chapter 11: for anything a company will build more than once, what you give AI is worth more than what AI gives you. They mark the page as a draft. Three of the four sentences are their best guess at what Priya believes, and they will not know until she says so. Agree. First with each other. In the thesis session from Chapter 8, Daniel attacks Maya’s line about accuracy, and they agree on what they will both say: it will be as right as the definitions it is given, and they will measure it together. Then with Priya. Maya opens the meeting by laying down the four sentences and asking which ones she does not buy. Priya pushes on the third at once: “That may hold for a software company, but our legal team has approved exactly one AI model and we will be on it for years.” Daniel writes it down word for word. It is the most useful thing anyone says that month, because it tells them where her world differs from the general case. Build. They build up from what she has agreed. Daniel asks the question from Chapter 9: “When a better AI model arrives, what in your plan would have to be rebuilt?” Priya answers it herself. “Most of it. We wrote the 178

Page 181 of 214. The Framework at Work prompts around one vendor.” And what would carry over? “The data. The definitions of a claim, a policy, a loss.” She has stated the implication. The decision follows: write down the definitions first, keep them where any approved AI model can use them, and start with the questions that already have a known right answer. Only then does a product appear, and the deci‐ sion criteria are written in her words: whatever they buy has to keep the defi‐ nitions outside the AI model. The trial is planned around the thesis, not around a tour of features. For two weeks it will answer claims questions with a known right answer, and both sides will count how many it gets right. Defend. The chief financial officer joins the second meeting and chal‐ lenges on scope: “We are not a software company.” Maya agrees that the fifth thesis needs a boundary, and they redraw it together: true for anything this company rebuilds every year, like its quarterly loss analysis. On Priya’s objec‐ tion about the single AI model they agree a timing signal instead of a date: when legal approves a second AI model, the plan’s first step pays off twice. Every challenge goes into the ledger, with its kind and its outcome. How it ends. Two endings are honest here, and the second teaches more. In the first, the trial lands, and the definitions project becomes the first phase of the contract. In the second, Priya buys from a competitor whose product also keeps the definitions outside the AI model. Maya and Daniel lose the deal and keep the account: the thesis Priya agreed to is still standing, written in her words, and when the next decision comes, it starts from there. That is what Chapter 10 meant by defending the thesis and the product separately. Where each step falls in a deal The six steps are not a separate sales process. They sit inside the deal you already run. 179

Page 182 of 214. Thesis-Based Selling round one BEFORE THE CALL disagreement is discovery DISCOVERY FIRST MEETING agree: lay the theses down defend, with JOLT PROPOSAL TO DECISION SOLUTION AND TRIAL build; the trial tests the signal AFTER SIGNING recheck at renewal Figure 2 4 The six steps inside the deal you already run. Before the first call you run round one on your own, and the inside half of Agree with the person who sells beside you. The first meeting opens with the theses on the table. Discovery is where the disagreements come, and each one is a note on what this buyer believes. The solution and the trial are Build, and the trial measures the signal the thesis named, not a tour of features. From proposal to decision is Defend, with JOLT’s moves if the buyer freezes. After the signature, recheck the thesis at every renewal, and carry the ledger to the next account. The six steps, one card each Each card says what the step is for, what you need, what you should have at the end, the questions to ask, how to know you are done, and the mistakes to avoid. Round one, step 1: List Purpose. Get every reason for and against the decision on paper be‐ fore sorting, so the beliefs underneath can surface. You need. One live deal, the decision in the customer’s words, your call notes, the language of any RFP, and what the buyer’s executives have said in public. You end with. The decision in one sentence, with no product in it, and ten to fifteen unsorted reasons, including the reasons someone gives for waiting. 180

Page 183 of 214. The Framework at Work Ask. What decision is the customer facing, in their words, and by when? What would they do if we did not exist? Which reasons have we heard them say, and which have we only said to them? What rea‐ son does someone give for not acting? Who in the buying group would write a different list? Done when. The decision has no product name in it, there are at least ten reasons, at least two are reasons to wait, and some are in the buy‐ er’s own words. Mistakes. Listing your features as reasons. Writing the decision as “choose vendor X”. Sorting while you list. Listing only what the champion says. Round one, step 2: Ask why Purpose. Find the few beliefs the reasons rest on. You need. The list. You end with. Three to five candidate beliefs, each with the reasons that hang from it. Ask. Why does that matter, until the answer is “because that is how the world is”? If this were false, would the decision still stand? Does it settle more than one decision? Which reasons share a root? Is this the buyer’s belief, or the one our product was built on? Done when. Ten or fifteen reasons have come down to three to five roots, each one carries weight in the decision, and you can say which reasons hang from which root. Mistakes. Stopping at a claim (“it is faster”). Landing on a platitude nobody would dispute. Treating a fear (“our competitors are doing it”) as a belief. Finding your own product’s thesis and calling it theirs. Round one, step 3: Write it down Purpose. Turn each belief into a sentence that can be agreed with, at‐ tacked and carried by someone else. 181

Page 184 of 214. Thesis-Based Selling You need. The candidate beliefs. You end with. Three to five theses in order, each passing the six checks in Chapter 5, each with what would prove it wrong and a first “then”. Marked as a draft until the buyer has responded. Ask. Who in the room would say no to this? Would it still be true if a competitor won? I would stop believing this if what? If it holds, what changes in their plan? Does each stand on the one before? Done when. A Thesis Sheet is filled in for each, and someone outside the deal can repeat each one after hearing it once. Mistakes. A product name creeping in. Hedging until it is a slogan. Tying it to one AI model release. Eight theses, which means some are still claims or duplicates. Round two, step 1: Agree Purpose. One voice inside your team, then a shared foundation with the buyer, with the disagreements found early. You need. The thesis page, your counterpart, and the map of the buy‐ ing group you already keep. You end with. A page signed by both of you with the unknowns list‐ ed. For each thesis, the buyer’s response: agreed, disputed and an‐ swered, or disputed and right. The first disagreement, word for word. Who in the buying group has heard which thesis. Ask. Inside: is this true on real data, why would a chief financial offi‐ cer care, and what do we not know? With the buyer: which of these do you not buy? Who in your company would push back on this one? If this held, what would you stop doing? Is this how your lead‐ ership sees it, or how you see it? Done when. Both names are on the page, the buyer has disputed or bounded at least one thesis, and you know whether the economic buyer holds the foundation thesis. Mistakes. A messaging document instead of an argument. Correcting each other in front of the customer. Reading politeness 182

Page 185 of 214. The Framework at Work as agreement. Agreeing only with the champion. Round two, step 2: Build Purpose. Work out the implication, then the decision, then the prod‐ uct, in steps small enough that the buyer says them first. You need. The agreed theses, and the constraints you learned from the disagreements. You end with. One or two written stacks. Decision criteria in the buy‐ er’s words that follow from the decision layer. A trial plan whose measure is the signal the thesis named. Ask. If this is true, what has to change in your plan? When a better AI model arrives, what would have to be rebuilt? What would carry over? What would you decide first? What would you need to see in the first two weeks to trust it? Done when. The buyer said at least one layer before you did, your champion can recite the stack from memory, and the product ap‐ pears only at the top. The decision layer would also admit a competi‐ tor’s product. If only yours fits, it is a pitch. Mistakes. Jumping from thesis straight to product. Steps too big for the buyer to say back. A demo before agreement. Decision criteria written so that only you pass. Round two, step 3: Defend Purpose. Let the thesis meet real pushback, concede what is right, keep what survives, and record it. You need. Theses with their falsifiers, and the ledger. You end with. Ledger lines (thesis, date, who, argument, kind, out‐ come), revised theses, an agreed signal and a date to look again for every timing challenge, and a decision to qualify out where the buyer is right that the foundation does not hold. Ask. Is this a challenge on fact, scope or timing? Here is what would change my mind; have you seen it? What is different about you? 183

Page 186 of 214. Thesis-Based Selling What would you need to see for “not yet” to become “now”? And to yourself: am I defending the thesis, or the product? Done when. Every challenge is logged, the falsifier was said aloud without being asked, each timing objection has a named signal, and the buyer has heard what you changed because of them. Mistakes. Treating it as objection handling, something to make go away. Bending the thesis to save the deal. Never conceding. Not logging. The Thesis Check MEDDIC tells you whether a deal is real. The Thesis Check tells you whether the ground under it is. Run both. Score each line 0 for no, 1 for partly, and 2 for yes with something written to show for it: a page, a quote, an email. 184

Page 187 of 214. The Framework at Work No. Check What a 2 looks like 1 Written The theses this deal rests on are written down, three to five, and pass the six checks 2 One voice Seller and engineer have signed the same page, with the unknowns listed 3 Their theses known We have written what each key role believes under this decision, not only what we believe 4 Match Our product follows from their theses, not against them 5 Tested The buyer has responded, and at least one disagreement is recorded word for word 6 Economic buyer The economic buyer has stated the foundation thesis in their own words 7 Travels The champion can recite the stack without us 8 Product follows Decision criteria were written from the decision layer, in the buyer’s words 9 Falsifier and signal Both sides agree what would prove the thesis wrong, and what the trial will measure 10 The other thesis We have written the thesis the main alternative rests on, including doing nothing Score Figure 2 5 The Thesis Check: ten lines, each scored 0, 1 or 2. There is no validated pass mark, and I will not invent one. Set your own after a quarter of use. A reasonable place to start: a deal that scores 0 on lines 5, 6 or 9 is not a commit, whatever the rest says. Some signs are worth a conversation whatever the score: the buyer agreed with every thesis; only the champion has heard them; a thesis names your product; there is no falsifier; or your product’s thesis contradicts the buyer’s. That last one no seller skill fixes. Either the conversation becomes the thesis, or the deal is out. For a team that runs its pipeline in a CRM, three fields are enough: the foundation thesis, whether and when the economic buyer agreed to it, and the agreed signal. 185

Page 188 of 214. Thesis-Based Selling For the manager A method lives or dies in the deal review. Five questions do most of the work. 1 “Read me the thesis this deal rests on.” 2 “Who on their side has disagreed with it, and what exactly did they say?” 3 “Has the economic buyer said it in their own words? When?” 4 “What would prove it wrong, and does the buyer know you said so?” 5 “If we lose this to a competitor, is the thesis still standing in the account?” Two habits keep it running. A thesis session, forty-five minutes, before the account plan on every new strategic deal. And a monthly thesis review, where the team pools its ledger lines and decides together which theses held, which got a boundary, which were revised and which were retired. Over time the team’s ledger becomes the shared record Chapter 10 describes, at the scale of a team. Coach two behaviors first, because both can be heard on a recorded call: saying the falsifier aloud, and writing down the first disagreement word for word. Introduce the steps one at a time, starting with the falsifier, the smallest new habit. Practice on the machine first, as Chapter 11 describes, then with a friendly customer. Watch for a ledger where every outcome is “held”, which means nobody is testing, and for theses copied from marketing or from Chapter 7. A copied thesis folds at the second question. To know whether the method is being run, track the leading signs: the share of strategic deals with a signed thesis page, the share of first meetings that open with the theses, the disagreements recorded per deal, how often the economic buyer’s agreement is recorded, and the 186

Page 189 of 214. The Framework at Work mix of ledger outcomes. To know whether it helps, compare no-deci‐ sion rates and cycle length for deals that scored high and low on the Thesis Check, after two quarters, against your own baseline. Run that comparison yourself. I have not run it for you, and I will not claim a result I have not seen. The lines to say These are the lines from the chapters that are worth having ready. The opener. “Before I show you anything, here are three things we believe are true about where this is going. I would like to know which ones you do not buy.” The falsifier. “Here is what would change my mind. Have you seen it?” Scope. “What is different about you?” Timing. “What would you need to see for ‘yet’ to arrive?” The concession. “You are right, that thesis does not hold for a com‐ pany in your position.” The build question. “When a better AI model arrives, what in your plan would have to be rebuilt?” The answer. “Nobody has done this long enough to prove it. Here is what we believe is true. Here is what would prove us wrong. Here is what follows for you if we are right. Tell me where you disagree.” That last one is the whole framework in five sentences. If you remember one line from this book, make it that one. 187

Page 190 of 214.

Page 191 of 214. Workbook Three Steps, Twice This is the Thesis-Based Selling framework as a practice. The two rounds from the introduction have three steps each, and each step has a page to write on. Run it once on a decision of your own before you run it on a customer. 189

Page 192 of 214. Thesis-Based Selling No. Step What you do Round one · Establish the thesis 1 List Write the decision and every reason behind it 2 Ask why Under each reason, ask why it matters until you reach a belief 3 Write it down One sentence for each belief, and what would prove it wrong Round two · Sell on the thesis 1 Agree Your own team first, then the buyer 2 Build If this is true, then: thesis, implication, decision, product 3 Defend Take the pushback, concede what is right, keep what survives Figure 2 6 The workbook: two rounds of three steps, each with a page to write on. Start at round one, step 1, with a live deal. Round one you can do alone. Round two starts with the person who sells beside you and then moves to the customer. 190

Page 193 of 214. Three Steps, Twice R O U N D O N E , ST E P 1 O F 3 List Pick one live deal. Write down the decision your customer is facing, in their words, with no product in it. Then list every reason you have heard for it, and every reason for waiting, in any order. THE DECISION REASONS TO ACT R E A S O N S T O WA I T 191

Page 194 of 214. Thesis-Based Selling R O U N D O N E , ST E P 2 O F 3 Ask why Take your three strongest reasons. For each one ask why it matters, then ask again, until the answer is a belief about how the world is. Expect several reasons to end at the same belief. THE REASON W H Y D O E S T H AT AND WHY DOES THE BELIEF M AT T E R ? T H AT ? U N D E R N E AT H 192

Page 195 of 214. Three Steps, Twice R O U N D O N E , ST E P 3 O F 3 Write it down Write each belief as one sentence, and next to it what would prove it wrong. Run the six checks from Chapter 5 on each. Keep three to five, in the order they build. N O. THE THESIS, IN ONE I WOULD STOP BELIEVING PA S S E S SENTENCE THIS IF CHECK 1 1 2 3 4 5 THE SIX CHECKS 1 Someone can disagree with it. 2 It is true no matter who wins the deal. 3 4 5 6 It names what would prove it wrong. Something follows from it. A tired executive can repeat it. It has a horizon and a signal to watch. 193 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6

Page 196 of 214. Thesis-Based Selling R O U N D T WO, ST E P 1 O F 3 Agree Your own team first, then the buyer. FIRST Your team Run the thesis session from Chapter 8 with the person who sells beside you: the account executive if you are the engineer, the engineer if you are the account executive. Each writes a list alone. Then attack each other’s, and keep what survives. W H AT S U R V I V E D B O T H MY LIST THEIR LIST AT TAC K S W H AT W E D O N O T K N O W Y E T , A N D W I L L S AY S O THEN The buyer Open the meeting with the theses and one question: which of these do you not buy? Write down the first disagreement word for word. It shows where their world differs from the general case. T H E O N E T H E Y D I D N OT B U Y, I N T H E I R WO R D S W H AT T H AT T E L L S M E A B O U T T H E I R W O R L D 194

Page 197 of 214. Three Steps, Twice R O U N D T WO, ST E P 2 O F 3 Build Fill this in from the bottom. Each layer should be a step small enough that the other person could say it before you do. The product goes in last. Beside each layer, note who said it first: them or us. PRODUCT what to buy DECISION what to do first I M P L I C AT I O N what it means for this company THESIS a claim about the world 195

Page 198 of 214. Thesis-Based Selling R O U N D T WO, ST E P 3 O F 3 Defend Keep one line for every challenge. Sort each as fact, scope or timing, and record the outcome: held, held with a boundary, revised, or retired. N O. DAT E WHO THEIR CHALLENGED ARGUMENT 196 KIND OUTCOME

Page 199 of 214. Three Steps, Twice K E E P The Thesis Sheet One page for each thesis that survives. Copy this page and keep the set where your team can read it. THE THESIS, IN ONE SENTENCE WHY I HOLD IT W H AT W O U L D P R O V E I T W R O N G I F I T H O L D S , T H E N ( T H E I M P L I C AT I O N ) THEN (THE DECISION) W H O H A S AT TA C K E D I T , A N D W H AT C H A N G E D S I G N E D , A N D T H E D AT E 197

Page 200 of 214. Thesis-Based Selling K E E P My Five Bets, Dated Write the date first. Then five bets on where your industry and AI will be. Keep the page and reread it in a year. T O D AY ’ S D AT E BET 1, AND THE YEAR I EXPECT IT BET 2, AND THE YEAR I EXPECT IT BET 3, AND THE YEAR I EXPECT IT BET 4, AND THE YEAR I EXPECT IT BET 5, AND THE YEAR I EXPECT IT REREAD ON 198

Page 201 of 214. Appendix The Field Notes The numbered notes from the chapters, in order, for when you want the argument on one page. Nº 01 A license is sold once. A seat is sold again at renewal. Consumption is sold every day the customer uses it. Nº 02 Features are compared inside a thesis. Whoever sets the thesis decides what gets compared. Nº 03 A reference is someone else’s result, borrowed. A buyer who cannot test your claim asks who already did. Nº 04 Nº 05 A first principle is private until someone else agrees with it. A thesis is a first principle you put on the table. The frameworks tell you how to run the sale. The thesis is what you stand on while you run it. 199

Page 202 of 214. Thesis-Based Selling Nº 06 Nº 07 A thesis is a claim someone in the room could say no to. If everyone would nod, keep sharpening it. A claim settles one comparison. A thesis settles the next ten decisions. Nº 08 The customer will find the gap between your seller and your engineer. Find it first, in a room where it costs nothing. Nº 09 A demo stays in the meeting. An argument the buyer can repeat goes everywhere the decision goes. Nº 10 Nº 11 Say the falsifier before the buyer asks for it. It is the cheapest credibility in selling. Let the machine attack your thesis. Never let it write one for you. 200

Page 203 of 214. Appendix List of Figures 1 Two rounds of three steps: every defense sends you back to a better thesis. 19 2 Features are compared inside a thesis, never across two. 39 3 The thesis decides which features count, before any product is shown. 47 4 Three answers to the buyer’s question, and none of them is a reference. 55 5 The Hype Cycle: expectations over time. Many buyers now stand in the trough. 58 6 Private principles give two conclusions. A thesis gives the room one. 62 7 Sales methods by era: what each supplied, and what each assumes. 71 8 The frameworks keep their jobs. What they stand on changes. 79 9 Three traditions, one working definition. 90 10 Three layers. The sale stands only where the two circles meet. 95 11 Keep asking until the answer stops being a claim: that is the thesis. 108 12 Worked example 1: A week at the beach 110 13 Worked example 2: Buying a database 111 201

Page 204 of 214. Thesis-Based Selling 14 Worked example 3: Leading a company through AI 112 15 The third thesis, drawn without a scale. Past the crossover people steer it. 122 16 Five theses, each one the consequence of the one above it. 126 17 People hear a change through the chain. Agents read it on one page. 128 18 Two instincts go into the page. One voice comes out of it. 137 19 The buyer agrees to the bottom layer first. The price shows up last. 145 20 Three things can happen with each thesis. Each tells you where to go next. 150 21 Sort the challenge first. Each kind has its own answer. 159 22 One line for every challenge, with its kind and its outcome. 161 23 The machine lands a fair hit. The thesis survives with a boundary. 170 24 The six steps inside the deal you already run. 180 25 The Thesis Check: ten lines, each scored 0, 1 or 2. 185 26 The workbook: two rounds of three steps, each with a page to write on. 190 202

Page 205 of 214. Resources Further Reading and Exercises Further reading These are the books this one leans on most, or that carry its argument further. Each is worth a seller’s time. The sales methods Neil Rackham, SPIN Selling, 1988. Research on thousands of sales calls, and the questions that move a large sale. Chapter 4 places it in its era. Michael Bosworth, Solution Selling, 1994. Selling by finding the buy‐ er’s pain and building the answer with them. Matthew Dixon and Brent Adamson, The Challenger Sale, 2011. The case for teaching the buyer something new, the closest of the methods to a thesis. Chapter 4 says where the two part. 203

Page 206 of 214. Thesis-Based Selling April Dunford, Sales Pitch, 2023. How to build a pitch on a market insight and a fair comparison of the alternatives, so that the buyer can decide with confidence. Andy Raskin, “The Greatest Sales Deck I’ve Ever Seen,” Medium, 2016. The strategic narrative: open with a big change in the world, not the product. John Care and Aron Bohlig, Mastering Technical Sales: The Sales Engineer’s Handbook, 2002. The craft of the sales engineer, from dis‐ covery to the demo. How markets adopt new technology Geoffrey Moore, Crossing the Chasm, 1991. Why mainstream buyers wait for references, and what that costs a new technology. Clayton Christensen, The Innovator’s Dilemma, 1997. How good companies lose by serving their best customers well. Thomas Kuhn, The Structure of Scientific Revolutions, 1962. Paradigms, and what happens when the ground under a field gives way. Deciding under uncertainty Annie Duke, Thinking in Bets, 2018. Judging a decision by its rea‐ soning, not its outcome. The habit behind your dated bets. Philip Tetlock and Dan Gardner, Superforecasting, 2015. What peo‐ ple who predict well do differently, starting with writing their predic‐ tions down. Robert Cialdini, Influence, 1984. Social proof and the other short‐ cuts buyers use, including the reference customer. AI and work Ajay Agrawal, Joshua Gans and Avi Goldfarb, Prediction Machines, 2018. AI as cheap prediction, and what becomes valuable when pre‐ diction is cheap. Ethan Mollick, Co-Intelligence, 2024. A practical account of working beside AI, from someone who uses it every day. 204

Page 207 of 214. Further Reading and Exercises Exercises Every chapter ends with steps for the coming week. Here they are in one place, with the workbook pages that go with them. Before you start. Write five dated bets on where your industry and AI will be in two, five and ten years. Workbook: My Five Bets, Dated. Chapter 1. Draw the path your last three deals took, date your last proof of concept, and name the unit your customer pays for. Chapter 2. Find the thesis under your product and under your com‐ petitor’s, and listen for the buyer’s. Chapter 3. Date and sort your customer stories, and write one belief that would outlive your company. Chapter 4. Name the method your team runs, find what it leans on, and check how much of that exists for AI. Chapter 5. Write three to five theses and run the six checks on them. Chapter 6. Round one on a personal decision, then on a live deal: list, ask why, write it down. Workbook: round one, steps 1 to 3. Chapter 7. Attack two of my five theses, and write one of your own. Chapter 8. Run a thesis session with the person who sells beside you. Workbook: round two, step 1. Chapter 9. Build one stack from thesis to product, and offer a thesis memo where an RFP is coming. Workbook: round two, step 2. Chapter 10. Keep a ledger of every challenge. Workbook: round two, step 3. Chapter 11. Run the four prompts on your best thesis, and log the machine’s attack. For every thesis that survives. Workbook: The Thesis Sheet. For every deal review. Field Guide: The Thesis Check, and the ques‐ tions for the manager. 205

Page 208 of 214.

Page 209 of 214. Notes Sources by Chapter Introduction. Geoffrey Moore, Crossing the Chasm, 1991. Michele Ballerini and others, “Interaction ruling animal collective behavior depends on topological rather than metric distance: evidence from a field study,” Proceedings of the National Academy of Sciences, 2008, on the starling flocks over Rome. John Care and Aron Bohlig, Mastering Technical Sales: The Sales Engineer’s Handbook, 2002. Robert Riefstahl, Demonstrating to Win!, 2011. David Maister, Charles Green and Robert Galford, The Trusted Advisor, 2000. Thomas Kuhn, The Structure of Scientific Revolutions, 1962. METR, “Measuring AI Ability to Complete Long Tasks,” March 2025. Blockbuster and Netflix: Marc Randolph, That Will Never Work, 2019, on the 2000 meeting; Blockbuster’s Chapter 11 filing, September 23, 2010. The consumerization of IT: the term used by Douglas Neal and John Taylor of the Leading Edge Forum from about 2001, and the Forum’s first paper on it, June 2004. Kodak: Steven Sasson’s 1975 prototype, as described by Kodak and in Sasson’s own accounts; Eastman Kodak’s Chapter 11 filing, January 19, 2012. Borders: the agreement for Amazon to run Borders’ online store, April 2001, and Borders’ Chapter 11 filing, February 16, 2011, both reported in The New York Times. Chapter 1. Epigraph: the standard caution in investment fund materials. Xerox 914, introduced 1959 and leased at $95 a month with 2,000 copies included (Fortune, “The office copier turns 50,” January 21, 2010). Xerox Professional Selling Skills and Xerox Learning 207

Page 210 of 214. Thesis-Based Selling Systems (Time, “Business: Xerox U”). Xerox’s training center in Leesburg, Virginia, opened 1974. Michael Bosworth, Solution Selling, 1994. Oracle’s up-front licensing, restatements and 1990 layoffs: Mike Wilson, The Difference Between God and Larry Ellison, 1997. Siebel Systems, full-year 2004 results, Form 8-K, February 2005. The “No Software” protest of February 2000: Marc Benioff and Carlye Adler, Behind the Cloud, 2009. Salesforce’s churn in 2005 and the birth of customer success: Nick Mehta, Dan Steinman and Lincoln Murphy, Customer Success, 2016. Amazon S3, launched March 14, 2006, and EC2, August 25, 2006 (AWS News Blog). Andrew Straus, “Sales Compensation in a Consumption Pricing World,” Snowflake blog, March 24, 2022. Snowflake’s forecast cut of May 24, 2023, and the share drop the next day (BNN Bloomberg, May 25, 2023). Intercom’s pricing of its Fin AI agent at $0.99 per resolution, announced 2023 (Intercom). Satya Nadella on the BG2 podcast with Brad Gerstner and Bill Gurley, December 2024. Sundar Pichai, Alphabet third-quarter 2024 earnings call, October 29, 2024. Andrej Karpathy on X, February 2, 2025. Collins Dictionary, Word of the Year 2025, November 2025. The software stock sell-off of early February 2026 (Semafor, “Software stock slide worsens on AI fears,” February 3, 2026). Salesforce Agentforce at $2 per conversation, Dreamforce, September 2024. Gartner, survey of 632 B2B buyers, press release of June 25, 2025. Chapter 2. Epigraph: Simon Sinek, Start With Why, 2009, and his TED talk “How Great Leaders Inspire Action,” 2009. The end of selective availability on the civilian GPS signal, May 1, 2000 (White House statement). Bill Gates, “The Internet Tidal Wave,” Microsoft memo, May 26, 1995. Steve Ballmer on the iPhone, interview with CNBC, January 2007. Steve Jobs, Macworld keynote introducing the iPhone, January 9, 2007; the first Android phone, the TMobile G1, on sale October 2008. Linus Torvalds’s announcement of Linux, comp.os.minix, August 25, 1991. SpaceX: first booster landing December 21, 2015; first crewed flight May 30, 2020 (NASA). Elon Musk, “The Secret Tesla Motors Master Plan (just between you and me),” Tesla blog, August 2, 2006. Steve Yegge, “Stevey’s Google Platforms Rant,” October 2011, on Jeff Bezos’s service-interface order at Amazon of about 2002. Apache Iceberg: created at Netflix and donated to the Apache Software Foundation (Apache Iceberg project history). Marc Andreessen, “Why Software Is Eating the World,” The Wall Street Journal, August 20, 2011. Offshoring and the Y2K work in India: Thomas Friedman, The World Is Flat, 2005. Ignaz Semmelweis, The Etiology, Concept and Prophylaxis of Childbed Fever, 1861; Sherwin Nuland, The Doctors’ Plague, 2003. Thesis: from the Greek thesis, a placing or setting down, from tithenai, to put or to place (Oxford English Dictionary). Chapter 3. Epigraph and the battery example: Elon Musk, interview with Kevin Rose, Foundation, 2012. Aristotle, Metaphysics, Book V, 1013a, on first principles. Robert Cialdini, Influence, 1984. Stanford HAI, AI Index Report 2025, on the cost of inference. MIT NANDA, The GenAI Divide: State of AI in Business 2025, and the criticism of its methods in the press that followed. Abraham Wald, Statistical Research Group, Columbia University, memoranda on aircraft survivability, 1943; Marc Mangel and Francisco Samaniego, “Abraham Wald’s Work 208

Page 211 of 214. Sources by Chapter on Aircraft Survivability,” Journal of the American Statistical Association, 1984. Gartner, Magic Quadrant research methodology; Forrester, The Forrester Wave methodology; IDC MarketScape methodology. Jackie Fenn, “When to Leap on the Hype Cycle,” Gartner, 1995. Gartner, Hype Cycle for Emerging Technologies, 2023, and Hype Cycle for Artificial Intelligence, 2025. John F. Kennedy, address to Congress, May 25, 1961; Alan Shepard’s flight, May 5, 1961; the Mercury, Gemini and Apollo programs (NASA history). George Akerlof, “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism,” Quarterly Journal of Economics, 1970, on asymmetric information. Gartner’s B2B buying journey research (Gartner for Sales): buying groups of six to ten people, and about 17 percent of buying time spent with suppliers. Chapter 4. Epigraph: George Santayana, The Life of Reason, volume 1, Reason in Common Sense, 1905. BANT: IBM’s Business Partner “BANT Criteria” forms of the 2000s. Mack Hanan, with James Cribbin and Herman Heiser, Consultative Selling, American Management Association, 1970. Robert Miller and Stephen Heiman, with Tad Tuleja, Strategic Selling, 1985. Michael Bosworth, Solution Selling, 1994. Neil Rackham, SPIN Selling, 1988; the 17 percent figure and “not the four little questions” from Huthwaite International’s account of the research. MEDDIC at PTC, 1996: Dick Dunkel and Jack Napoli, as told on Masters of MEDDICC; John McMahon, The Qualified Sales Leader, 2021; Andy Whyte, MEDDICC, 2020. Matthew Dixon and Brent Adamson, The Challenger Sale, 2011, on commercial teaching (teach, tailor, take control, and the six-step teaching choreography); Noel Capon, “Revisiting The Challenger Sale,” Velocity, 2015. Brent Adamson, Matthew Dixon, Pat Spenner and Nick Toman, The Challenger Customer, 2015. Keenan, Gap Selling, 2018. Matthew Dixon and Ted McKenna, The JOLT Effect, 2022. Andy Raskin, “The Greatest Sales Deck I’ve Ever Seen,” Medium, 2016, on Zuora’s deck and the strategic narrative. April Dunford, Obviously Awesome, 2019, and Sales Pitch: How to Craft a Story to Stand Out and Win, 2023. Chapter 5. Epigraph: attributed to Martin Luther at the Diet of Worms, 1521, from early printed accounts of his reply. Martin Luther, Disputation on the Power and Efficacy of Indulgences (the Ninety-five Theses), October 31, 1517; on Tetzel and on the door, the debate summarized in Lyndal Roper, Martin Luther: Renegade and Prophet, 2016. Union Square Ventures: Brad Burnham’s one-line thesis, 2011; “Thesis 2.0,” December 2015; “Thesis 3.0,” April 2018 (usv.com). Karl Popper, The Logic of Scientific Discovery, 1959. Jeff Bezos, fireside chat with Werner Vogels, AWS re:Invent, November 29, 2012. Giuseppe Peano, Arithmetices principia, nova methodo exposita, 1889, which began from one. Alfred North Whitehead and Bertrand Russell, Principia Mathematica, volume I, 1910 (54.43), and volume II, 1912 (110.643). Nikolai Lobachevsky (1829), János Bolyai (1832) and Bernhard Riemann (1854) on geometries without Euclid’s fifth postulate. The Declaration of Independence, 1776, second paragraph. Chapter 6. Epigraph: Plato, Apology, 38a. The five whys: a practice of Sakichi Toyoda, made part of the Toyota Production System by Taiichi Ohno, Toyota Production System, English edition 1988. 209

Page 212 of 214. Thesis-Based Selling Chapter 7. Epigraph: Peter Thiel with Blake Masters, Zero to One, 2014. Chapter 8. Epigraph: Jeff Bezos, letter to Amazon shareholders, 2016, on “disagree and commit.” Chapter 9. Epigraph: attributed to Archimedes, as reported by Pappus of Alexandria, Collection, Book 8. Euclid, Elements, about 300 BC. Stipulations of fact before trial: United States Federal Rules of Civil Procedure, Rule 16(c)(2)(C). Cunningham’s Law, coined by Steven McGeady in a reader comment on The New York Times‘s Schott’s Vocab blog, May 28, 2010. Gartner, as in Chapter 3. Chapter 10. Epigraph: John Stuart Mill, On Liberty, 1859, chapter 2. Charlie Munger, commencement address, USC Gould School of Law, 2007. Richard Feynman, “Cargo Cult Science,” Caltech commencement address, 1974. Paul Saffo, “Strong Opinions, Weakly Held,” 2008. Chapter 11. Epigraph: Garry Kasparov, “The Chess Master and the Computer,” The New York Review of Books, February 11, 2010. Forrester, Buyers’ Journey Survey, 2024. Anil Doshi and Oliver Hauser, “Generative AI enhances individual creativity but reduces the collective diversity of novel content,” Science Advances, 2024. 210

Page 213 of 214. About The Author Fru Nde is an enterprise seller and sales engineer. He has spent his career inside some of the largest and most strategic accounts in the world, and he has contributed, directly and indirectly, to more than $100 million in sales. F R U N D E What he sells is data, AI and the change that comes with them. His ground is the complex enterprise deal, where a decision passes through many hands before anyone signs, and where how a company decides matters as much as what it buys. He has watched buying patterns shift and deal cycles stretch and shrink, and he has seen what wins in the trenches and what quietly falls apart. This book is what he has learned about selling when there is nobody to point at, in a time of radical change. ALSO BY THE AUTHOR Do The Opposite A field guide to the counter-intuitive rules of wealth, power, and reality. DTO. F R U. D E V Your Game Plan for Success with AI 5 simple steps to get organized, get ready and win big with AI. Y G PA I . F R U . D E V He writes at fru.dev. 211 Intent Engineering What is left for you to do when AI can do the work. IE.FRU.DEV

Page 214 of 214. THESIS-BASED SELLING BY FRU NDE FIRST EDITION, 2026 SET IN EB GARAMOND & INTER

Thesis-Based Selling

That is the whole book.