First Customer

Before you begin

Before You Begin

Most of my students are twenty-four. They are smart, they already use AI every day, and they are about to take their first real job. At the dinner table they are being told a story about that job that was true for their parents and is not true for them. This book is the other story.

It rests on one sentence, and the sentence is the whole strategy:

Let your first employer be your first customer.

I have been saying it to students for a while now. I have never written it down at length, and a sentence that short deserves to be taken apart word by word, because every word in it is doing a job. That is Part Two. Before it, in Part One, I want to show you where the thing called a job came from, because you cannot hold a job lightly until you understand that it is young, that it was assembled, and that most of what was stapled to it was stapled on for reasons that had nothing to do with you. After the sentence, Part Three is about money, Part Four is about your AI, and Part Five is about what the market does next.

How the book is built

Each chapter rests on a first principle: something that is true by structure, not by survey. Where I lean on a piece of history, I name the source once and move on. I have kept statistics out on purpose. You do not need a number to understand why a firm exists or why a mortgage is a sale of your freedom. You need the shape of the idea, and then you need to go and look at your own life.

Most chapters end with an assignment. The assignments are short, and they are not reflective exercises. Each one produces something: a number, a list, three questions you will put to a real company, a record you will start keeping. Do them in order. By the end you will have a floor, a stack, a log and a price, which is everything the strategy needs.

What you need

A laptop. An AI you already work with, from one of the major companies, that you can connect to your own email, calendar and files. A copy of whatever offer letter or employment agreement you have signed or are about to sign. A spreadsheet with one column in it. That is all.

If you have read Super Admin, the advanced course, you will recognize the stack chapters, and you can move through Part Four quickly. If you have not, Part Four stands on its own and tells you where to go deeper.

A note on the equation underneath

Everything I write sits on one framework, which I call the Reality Equation. Reality is a ratio: what actually arrives, divided by what you expected. You do not choose the expectation. You inherit it, unconsciously, from everything that has happened to you and around you. A twenty-four-year-old’s expectation of a job was built out of twenty years of dinner-table conversation with people who were twenty-four in a different world.

That matters here for one reason. Nobody can reach in and change what you expect, not you, not your parents, not me. The only thing that retrains an expectation is the record of what actually happens. So the strategy in this book is not a new way of thinking about work. It is a set of actions that leave a record. Deliver finished work to the company that hired you, keep your costs low enough that you could walk, keep your own AI and your own log, and in a year or two your expectation of what a job is will have moved on its own, because the Actuals moved it. You cannot talk yourself into this. You can only act your way into it. That is the whole reason the book is built out of assignments.

Chapter 1

A Job Was a Piece of Work

The word is younger than you think, and it did not mean what it means now.

In the middle of the sixteenth century, English speakers began saying “a jobbe of worke,” and they meant a task: a lump of work, a single thing to be done, as opposed to the continuous labor of a servant or a farmhand. The origin is uncertain; the dictionaries guess it may come from an older word for a lump. By the sixteen-twenties, “job” on its own meant a piece of work. By the sixteen-fifties it meant a piece of work done for pay. A printer in the seventeen-nineties did “job work”: handbills, posters, the small one-off pieces between the books.

The sense you grew up with, a job as a paid position of employment, a thing you have rather than a thing you do, is recorded only from the middle of the nineteenth century. For roughly three hundred years before that, a job was something you finished.

I want you to hold that, because it is the oldest meaning and it is the one this book is trying to give back to you. A job is a piece of work. You take it, you do it, it is done, you are paid. The permanent, continuous, identity-shaped thing your parents mean by the word is the newcomer.

Paid by the day

Go back a little further and look at how skilled people actually worked before the factory. In the craft trades of medieval and early modern Europe there were three ranks: apprentice, journeyman, master. The apprentice was a beginner bound to a master for a term of years. The master owned the shop. And the journeyman, the qualified worker in the middle, took his name from the Old French journee, a day: he was a man who worked by the day, for wages, for whoever would hire him. He was not an employee in the modern sense. He had a trade, he had tools, and he sold his days.

Below and around the guilds there was an even looser arrangement, which historians call the domestic system or the putting-out system. A merchant would put out raw material, wool or flax or iron, to families working in their own homes, with their own tools, on their own hours. The families turned it into finished goods, and the merchant paid them, often by the piece. Nobody in that system had a job in your parents’ sense. They had customers.

The factory took the tools

What changed was the factory. When weaving was mechanized in England in the late seventeen-hundreds, the machines were too large and too expensive for a cottage. They had to be gathered under one roof, and the roof belonged to whoever could afford the machines. So the workers came to the machines, and for the first time at scale, the person doing the work did not own the tools the work required. The employer owned the tools, owned the raw material, and set the hours.

Marx, writing in the eighteen-sixties, called this the historical process of divorcing the producer from the means of production, and whatever you think of the rest of his argument, the observation is true. The craftsman had owned his tools. The factory hand did not. And once you do not own the tools, you cannot sell finished work, because you cannot produce any on your own. You can only sell your time, in the building where the tools are, on the hours the owner sets.

That is the moment the job stopped being a piece of work and became a position. Not because anyone decided it should. Because the tools moved.

Figure 01

Where the tools are, and what gets sold

Tools in the worker’s handsTools in the owner’s building

1. Journeymansells days; owns his tools
2. Putting-outsells pieces; works at home
3. Factorysells hours; owner owns tools
4. Organization mansells a career; the bundle
5. Laptoptools fit in a bag again
6. Personal AIthe method travels too

What gets sold

Left: finished pieces, to customers. Right: time, to an employer. The arrangement follows the tools.

Why this matters to you

Here is the first principle of Part One, and the whole book turns on it. The shape of work follows the ownership of tools. When the tools are in the worker’s hands, work is sold as finished pieces to customers. When the tools are in the owner’s building, work is sold as time to an employer. Every arrangement in between is a mixture, and the mixture is set by where the tools are.

For two hundred years the tools were in the building. Then, slowly, they began to leave it, and Chapter 4 is about how far they have gone. For now, notice only this: the word you have been taught to want, the job, named a piece of work for three centuries before it named a position for a century and a half. The older meaning is not nostalgia. It is coming back, and the tools are the reason.

Chapter 2

Why Firms Exist

Here is a question that sounds too simple to ask. If markets are so good at organizing work, why do companies exist at all?

Think about what a company is. Inside it, work is not bought and sold. Nobody in accounting invoices the person in sales for a report. A manager says do this, and it is done, and no price is discovered and no contract is signed. A company is a patch of the economy where the market has been switched off and replaced with instructions. If the market were really the best way to organize everything, that patch would not be there.

A young British economist named Ronald Coase asked that question in 1937, in a paper called The Nature of the Firm, and his answer is the second first principle of this book.

The market has a cost

Using the market is not free. To buy a piece of work from outside, you have to find someone who can do it. You have to discover what it should cost. You have to describe what you want precisely enough that a stranger can deliver it. You have to negotiate, write it down, check the result when it comes back, and do all of that again for the next piece. Each of those steps costs time and attention and, often, money. Coase’s phrase was that there is a cost of using the price mechanism. The costs were later given the name they carry now: transaction costs.

A firm exists wherever those costs are higher than the cost of simply hiring someone and telling them what to do. Instead of buying each piece of work from the market, the company buys the person’s time, in bulk, and directs it. It gives up the market’s prices and gets, in exchange, the right to say do this without a new contract every morning.

And Coase went one step further, which is the step that matters to you. A firm grows until the cost of organizing one more piece of work inside it equals the cost of buying that piece from the market. The boundary of the firm, the line between what it makes and what it buys, is set by those two costs. Move either one, and the line moves.

That is the whole theory, and it was half of what won him the Nobel prize in economics in 1991, fifty-four years after he wrote it. The prize committee put the point in one sentence: if transaction costs were zero, no firms would arise.

You were a make-or-buy decision

Business people have a plainer name for the boundary question. They call it make or buy. Make the part or buy it. Run the kitchen or cater. Employ the accountant or hire the firm. Harvard’s business school was publishing studies with that title in the nineteen-forties, before the economists made it the test case for the theory.

Hiring you was a make-or-buy decision, and the company chose make. It looked at a stream of work it needed done, judged that the market could not be trusted to deliver that work reliably and cheaply enough, piece by piece, and decided to bring the capability inside by putting you on payroll. It bought your Tuesdays in bulk rather than buying your deliveries one at a time, because buying the deliveries one at a time looked too expensive. Too hard to find the right person. Too hard to describe each piece. Too hard to check.

Figure 02

You were a make-or-buy decision

Make

Put the capability inside. Hire a person, pay for their Tuesdays, own what they produce, carry the bundle.

Chosen when the market cannot be trusted to deliver the outcome cheaply and reliably.

Buy

Buy the outcome from outside. Pay a vendor for finished work, on a term, with a test.

Chosen when finding, specifying and checking the work costs less than owning the person who does it.

Hiring you was a Make. Every finished delivery you log makes the Buy real for the same company.

That is not an insult. It is a description of a cost. And it tells you exactly what the strategy in this book is doing underneath. Every chapter in Part Two is a way of lowering, for one company, the cost of buying your work from the market: making yourself easy to find (they already found you), making your deliveries easy to describe (finished, with a test), and making them easy to check (the log). When those costs fall far enough, the decision the company made when it hired you comes up for review, and it comes up on your terms.

The theory is leading, not settled

I should say plainly that Coase’s answer is the leading explanation for why firms exist, not the only one. Other economists have argued that firms exist because some work is done by teams whose output cannot be split into individual contributions, so someone has to watch; or because of who owns the machines; or because companies are better than markets at moving the kind of knowledge that cannot be written down. Each of those is a real argument with a real literature behind it, and I am not going to adjudicate them here.

I lean on Coase for a practical reason. His is the only one of the explanations that gives you, the person on the inside, something to do. Monitoring, asset ownership and tacit knowledge are the company’s problems. The cost of finding, describing and checking your work is something you can lower yourself, from your own chair, starting Monday. And in Part Five I will argue that the thing that has changed most about work in the last few years is exactly that cost.

Chapter 3

How the Bundle Accreted

Open a modern offer letter and look at what comes with the salary. Health insurance. Retirement contributions. A laptop. A title. An email address with the company’s name after yours. Training. A badge. Sometimes a parking space. The job your parents told you to want is not a wage. It is a bundle, and almost everything in the bundle is less than a century old.

I used to call this an accident. The research for this book taught me to be more careful. The bundle was not one accident. It accreted, the way a reef does, out of a series of small decisions made for reasons that had nothing to do with you. But the effect is the same as if it had been an accident, because nobody designed it, nobody chose it on your behalf, and nobody is responsible for keeping it.

How insurance got stapled to the paycheck

The clearest example is health insurance, and it is worth telling plainly, because it is the part of the bundle that most often keeps people from treating a job as a customer.

In the United States, group health coverage through employers was already growing in the late nineteen-thirties. Then the war came, and with it a problem: with millions of workers gone, employers competing for the rest would bid wages up and set off inflation. So in October 1942 Congress froze wages at their September levels. But the law that froze wages left a door open. It exempted insurance and pension benefits, within reason, from the freeze. Employers who could not offer more pay could offer more benefits, and the board that administered the freeze treated those benefits as noninflationary because they were not cash.

The next year, the tax authority issued a narrow ruling that an employer’s contributions to group health coverage were not taxable income to the worker. It was a limited ruling, and the rulings that followed it over the next decade were inconsistent. In 1954, when Congress rewrote the tax code, it settled the matter and made the exclusion uniform. From then on, a dollar of health insurance bought through your employer was worth more than a dollar of salary used to buy the same insurance yourself, because the first dollar was never taxed.

Figure 03

How the bundle accreted

1. 1942Wages frozenCongress freezes wages for the war and exempts insurance and pension benefits from the freeze.
2. 1943Benefits untaxedA narrow ruling: an employer’s payments for group health coverage are not the worker’s taxable income. Withholding from paychecks returns the same year.
3. 1954Made uniformThe rewritten tax code settles the exclusion. A dollar of insurance through the job is now worth more than a dollar of salary.
4. 1956The Organization ManA generation belongs to the company and expects to be taken care of in return.

Nobody in that story was thinking about whether a twenty-four-year-old in 2026 should be able to leave a job without losing her doctor. They were thinking about inflation in 1942 and tax uniformity in 1954. The staple was administrative. It is still holding.

The rest of the reef

The same decade gave the paycheck its modern form. Withholding tax from wages had been tried in the nineteen-teens and dropped; in 1943 it came back for good, and from then on the employer stood between you and the government, collecting your taxes before you saw your pay. A little over a decade later a writer at Fortune described the result in a book called The Organization Man: a generation that belonged to the company, took its identity from the company, and gave the company its loyalty in return.

That expectation is the one your parents inherited, and the one they are handing to you at the dinner table. It was true, partly, for one generation, in one country, during one unusual stretch of history when the tools were all in the building and the staples were all fresh. It is not a law of nature. It is a reef.

What the bundle costs you

Every item in the bundle does two things. It gives you something, and it makes leaving more expensive. The insurance is real, and it means you cannot leave without replacing it. The retirement match is real, and it vests on a schedule designed to keep you. The laptop is real, and everything you build on it belongs to someone else. The title is real, and it describes a seat rather than a capability. Each staple is a small reason to stay, and together they are the reason most people never have the conversation in Chapter 8.

A vendor does not take the bundle. A vendor carries their own insurance, their own retirement, their own tools and their own name, and prices all of it into the invoice. That sounds expensive, and it is, until you notice that the employee was paying for the bundle too, out of a salary that was lower than it looked by exactly the cost of the bundle. The company was never giving you the insurance. It was buying it on your behalf, with your money, and attaching a condition: stay.

The strategy in this book asks you to unbundle, deliberately, over a few years, so that the day you make the offer in Chapter 8, nothing in the bundle is holding you to the chair. Part Three is about how.

Chapter 4

The Means of Production Came Back

In Chapter 1 I said the shape of work follows the ownership of tools, and that for two hundred years the tools were in the owner’s building. This chapter is about how they left.

The knowledge between their ears

The first person to notice, as far as I can tell, was Peter Drucker. In the late nineteen-fifties he began using the phrase knowledge worker for people whose work was mostly thinking, and he spent the rest of his life working out what that meant. By 1999, in Management Challenges for the 21st Century, he had reduced it to a sentence: knowledge workers own the means of production. It is the knowledge between their ears, he wrote, and because they own it, they are mobile.

That was a reversal of the factory, and Drucker knew it. The factory worked because the tools were too large and too expensive to leave the building, so the worker had to come to them and sell his time. If the tool is knowledge, it leaves the building every evening in the worker’s head. The owner cannot lock it up. The owner can only make it attractive to come back.

Drucker was right, and he was early, because in 1999 knowledge still needed a building to do anything with. The analyst’s knowledge needed the company’s data, the company’s software, the company’s colleagues and the company’s desk. The means of production were between his ears, but the means of applying them were still in the owner’s hands.

The laptop

Then the tools shrank. Within a decade of Drucker’s sentence, most of what a knowledge worker needed to apply his knowledge fit in a bag: the computer, the software, the connection to everything. Companies noticed, and after a brief struggle they gave up pretending otherwise. Around 2010 a few large employers began letting people connect their own phones and laptops to company systems. The policy had a name almost at once, bring your own device, and within a few years most of the people who worked with information owned the devices they worked on.

Read that against Chapter 1. Bring your own device is the journeyman’s arrangement coming back. The worker arrives with his tools. The employer provides the situation and the work. The line between them is drawn by permissions rather than by walls.

The personal AI

And now the last tool has left the building.

The laptop carried the knowledge worker’s tools. It did not carry his method. How he actually did the work, what he had learned about the standard and the clocks and the gap, what he had tried that failed, how he liked work to come back to him: all of that still lived in his head and in the company’s systems, and when he left, the head went and the systems stayed.

A personal AI carries the method. Working beside you for a year, it learns how you work, what you know, where your records are, which decisions you settle in your sleep and which ones surprise you. It can take an outcome you describe and deliver finished work in your way, with your standards, on your clocks. And it lives on an account you own.

I have written about this at length in a piece called The Charge Is Yours, and I will not repeat the argument, except for the one conclusion that belongs here. The most valuable thing any AI holds is its model of one person, and that model is built mostly outside the job and compounds only if it travels. The company cannot own it, because the company does not see most of what builds it. The person can, and should.

So the tools have made the whole journey. From the journeyman’s hands, into the factory, into the organization, and now back out, first into a bag and then into an account. Drucker’s sentence is finally true in full. The means of production are yours, the means of applying them are yours, and the method of applying them is yours. The only thing the company still owns is the situation: its customers, its deadlines, its systems of record. That is a great deal to own, and it is exactly what a vendor needs a customer for.

Chapter 5

Let

The sentence begins with the smallest word in it, and the smallest word is the one people skip.

Let your first employer be your first customer. Not make. Not ask. Not negotiate. Let.

“Let” is permission, and the person granting it is you. Nobody at the company is going to hand you a customer relationship. There is no form for it. There is no title. The posture I am describing is something you take, quietly, on the first morning, before anyone has noticed you have taken it, and it does not require anyone’s agreement to be real. You can hold a W-2, sit in the chair they gave you, log in to the laptop they issued, and still be, in every way that matters to how you work, a vendor delivering finished work to a customer who happens to pay you every other Friday.

That is the first thing “let” means. The posture is yours to adopt.

The second thing it means is softer. To let something be is to allow it, to not fight it. I am not telling you to march into your manager’s office and announce that you are now a contractor. I am telling you to let the company be what it already is to you, a buyer of your work, and to stop pretending it is something else. A company is not your family. It is not your identity. It is not the thing that will take care of you. Companies have been saying that for about seventy years, and they said it most loudly during the one stretch of history when it was briefly, partly true. You do not have to argue yourself out of that story. You work like a vendor, and the record does the arguing. The company is a customer, and customers are wonderful. You just have to serve them properly.

There is a third sense, and I like it, so I will give it to you even though it is British. To let a house is to rent it out. You keep the house. You let out the use of it. When you let your employer be your customer, that is exactly the arrangement. You are letting out your capacity, your hours and your attention, for a price, for a term. You keep the asset. The asset is you: your judgment, your method, your AI, your record of finished work. None of that is on the lease. If you forget that, if you let the asset go with the tenant, you have not let anything. You have sold.

Why the posture has to come first

Students always want to start with the money. How do I get them to pay me as a contractor? How much do I charge? Those are real questions, and Parts Three and Five answer them. But they are the wrong place to start, because the arrangement follows the posture and never the other way around.

A company cannot buy finished work from someone who has never delivered any. It cannot price an outcome it has never seen you produce. The conversation about converting a job into a customer relationship is only possible after a record exists, and the record only exists if you have been working like a vendor since the first week. So the order is fixed. Posture first. Record second. Money third. Students who reverse it end up negotiating for a status they have not earned, and companies, reasonably, say no.

What the posture is not

It is not detachment. A good vendor cares more about the customer’s outcome than most employees care about their boss’s, because the vendor’s next invoice depends on it. It is not coldness. The best vendors I have ever worked with were warm, present and generous with their attention, and they were generous because they had decided where their attention went instead of letting it be taken.

It is not insubordination. The customer sets the outcome. You own the path. If the customer says the report goes out Thursday, the report goes out Thursday. What changes is not who decides what; it is how the work comes back. It comes back finished.

And it is not a secret. You do not have to hide the posture, and you do not have to announce it. Just work that way. Within a few months, people will describe you as someone who gets things done, which is the ordinary English for a person who delivers finished work and does not hand it back half done with a question attached.

Chapter 6

Your First Employer

The second phrase tells you to take the job.

I want to be plain about this, because a book about treating your employer as a customer can be misread as a book about not having an employer. It is the opposite. I want you to take the job, and I want you to take it seriously, and here is why.

The hardest problem any vendor on earth has is finding the first customer. Not the second. The first. The first customer has to trust someone with no record. The first customer has to explain their business to a stranger, show them where the files are, teach them the names of things, forgive the early mistakes, and pay anyway. Almost nobody will do that for a twenty-four-year-old with a degree and an AI.

Except an employer. An employer will do all of it, and call it onboarding.

Think about what has happened by the time you sit down on your first day. They searched for you. They read your application. They interviewed you, twice, maybe three times. They checked your references. They argued about you in a room you were not in and decided, over other people, that they wanted you. Then they agreed to pay you before you had produced a single thing. From a vendor’s point of view, that is not a job. That is a customer acquired, at their expense, before you have delivered anything.

So the first employer is not an obstacle to the strategy. The first employer is the strategy’s first move.

The apprenticeship was always this

There is nothing new here. The oldest career path in the world is the one where you work inside someone else’s shop until you can run your own, and the shop you worked in becomes the first thing your own shop is measured against. Benjamin Franklin was indentured to his brother James as a printer’s apprentice in Boston as a boy. He learned the trade on his brother’s presses, with his brother’s customers, on his brother’s paper. He left before his time was up, which is a story for another day, and five years later he was a partner in his own print shop in Philadelphia, and soon after that its sole owner. The customers he served there were the kind of customers he had learned to serve in Boston. The shop in Boston was not a detour from Franklin’s career. It was where the career was manufactured.

An apprentice does not resent the master for owning the shop. The apprentice is there to learn what the shop knows: what finished work looks like in this trade, what the customer actually pays for, where the time goes, what breaks. You cannot learn those things from outside. You have to be inside, with permission, for a while.

What you are actually there to learn

Your first employer will teach you three things that no school and no AI can teach you, because they are local.

The first is the standard. Every trade has a picture of what done looks like, and the picture is never written down. In one company a finished analysis is two pages; in another it is a deck; in another it is a sentence in a meeting. You learn the standard by producing work and watching what gets accepted. It is worth more than the salary.

The second is the clocks. Every business runs on deadlines it rarely writes down: when the truck leaves, when the books close, when the client expects the call. I have called these the spoil clocks elsewhere, because they are the moments after which a piece of work is worth less. You cannot deliver finished work on time until you know when on time is, and the only way to learn that is to be inside the building when the clocks go off.

The third is the gap. Inside a company you will see, every week, work that the company needs done and cannot get done. Not because it is hard. Because it falls between people, or because nobody owns it, or because the one person who could do it is doing something else. That gap is where your first paid delivery as a vendor will come from, and you will only see it from inside.

My own first employer

My own first employer was a company I helped start. I was one of its first two employees. The money came from a group of large electronics companies and a venture firm, and the companies that put up the money were the companies the thing was being built for. So my first employer and my first customers were the same people. I did not plan it that way. I have come to think it is the natural shape of a first job, and that most people never notice that the person signing their paycheck is also the person buying their work.

Chapter 7

Be Your First Customer

Now the word that changes how you work: customer.

A customer and an owner buy different things from you. An owner buys your time. The owner wants you in the chair from nine to five, available, responsive, present. What you produce in that time is the owner’s, and so is the time itself, and if the owner wants you to spend Tuesday in a meeting about the meeting, that is the owner’s right, because the owner bought Tuesday.

A customer buys an outcome. The customer does not care where you were on Tuesday. The customer cares whether the thing is done, whether it is right, and whether it arrived when it was supposed to. The customer does not want to supervise you. Supervision is a cost to the customer, and a good vendor removes it.

Every job you will ever have is some mixture of these two purchases. The strategy is to move the mixture, from the first week, as far toward the customer side as the work allows. You do that by changing one thing: how the work comes back.

Finished, with a test

When an employee is given a task, the task tends to come back as progress. A draft. A first pass. A few options. A question about which direction to take. Each of these is the work returning to the person who assigned it, asking for more of their attention. Employees do this because school trained them to show their work, and because handing back something unfinished feels safer than committing to something finished.

A vendor cannot do this. Imagine hiring someone to repair your roof and having them come down the ladder at noon with three options and a question about which shingle you prefer. You hired them so that you would not have to think about shingles. You want to look up and see a roof.

So the rule is: the work comes back finished, and it comes back with a test. Not “here is a draft of the report,” but “the report is done; it says the vendor overcharged us by this amount in these three months, and here is how I checked it.” Not “I looked into the scheduling problem,” but “the schedule for next month is posted, every shift is covered, and here are the two people I had to move.” When you return work this way, you are returning an outcome. You are also, without saying so, teaching the company what it is actually buying from you, which is the thing it would have to buy from a vendor if you left.

The two words my students learn to type to their AI are the two words you should hope to hear from your customer: done, and accepted. A manager who can look at your work and say only “accepted” or “not yet” is a manager you have made into a customer. That is the whole move.

Small matters down, great matters up

A vendor posture has a second half, and it concerns what you bring to the customer rather than what you take from them.

An employee brings questions up. May I? Which one? What do you think? Each question is reasonable and each one is a small withdrawal from the manager’s attention, and a manager’s attention is the scarcest thing in the building. A vendor sends the small matters down, to their own AI and their own judgment, and brings up only the matters that are actually worthy of the customer’s attention: a real decision only the customer can make, a risk they need to know about, money, anything that goes out under the company’s name to someone outside.

I have written about this elsewhere as the great matters. The point here is practical. The fewer small matters you send up, the more your manager experiences you as a vendor and the less as a subordinate, and that experience is what the conversation in the next chapter depends on.

The log starts on day one

One more habit, and it is the one students most often skip, because its value is not visible for months.

Every time you deliver something finished, write it down. One line. The date, what was delivered, who accepted it, and what changed because of it. Not what you did. What was delivered. “Posted next month’s schedule, all shifts covered, accepted by Dana” is a line. “Worked on scheduling” is not.

This is your delivery log, and it is the single most important document in this book. It is the record a customer relationship is eventually built on. It is also, as Part Four will argue, the thing that is about to replace the resume. Start it in the first week. Keep it in a file you own, on a machine you own, not in the company’s systems, because it is yours and it goes with you.

Chapter 8

The Make-or-Buy Decision Runs Both Ways

In Chapter 2 I said that hiring you was a make-or-buy decision, and that the company chose make. Here is the part nobody tells you. That decision is not permanent, and it is not only theirs.

Once you have spent a year delivering finished work with a log, you have changed the facts the decision was based on. The company originally could not buy this outcome from the market because the market did not have anyone who knew their standard, their clocks and their gap. Now it does. You. You are the vendor they could not find, and you built yourself inside their building, on their dime, with their permission.

The conversation

Somewhere between the first year and the second, there is a conversation available to you that is not available to most employees. It is not a resignation and it is not a demand. It is an offer, and it sounds something like this.

“Here is what I have delivered in the last twelve months.” The log. “Here is what the company spends to have me deliver it.” Salary, benefits, the desk, the laptop, the share of a manager’s attention. “I would like to keep delivering it, and I would like to do it as your vendor rather than your employee. Same outcomes, same deadlines, priced as a monthly retainer. You stop paying for my Tuesdays and start paying for the thing you actually want.”

Some companies will say yes, and some of those will be relieved, because they never wanted to own your Tuesdays either. Some will say no because the work cannot be bought: it is regulated, or it is bound up in a team in a way that cannot be separated, or the company has rules about contractors that are older than the people enforcing them. And some will say no because they want the control more than they want the outcome. That last answer is not a failure. It is information, and it is information you could not have gotten any other way. You have learned what the company is actually paying for, and you can decide whether that is a customer you want to keep.

Two offer letters

I think the conversation will get easier every year, and I will say why in Part Five. For now, hold this picture. Within a few years, a capable person’s first offer will routinely arrive in two versions: one as an employee, one as a vendor on retainer, same work, different terms. Companies will do this because the people they most want to hire will ask for it, and because the cost of writing the second letter is almost nothing. When that happens, the strategy in this book stops being a strategy and becomes the normal shape of a first job.

You do not have to wait for it. You can hold the posture now, build the log now, and have the conversation when the log is thick enough. The only thing that can stop you is the thing the next three chapters are about: if you need the salary to live, you cannot make the offer, because an offer you cannot afford to have refused is not an offer. It is a request. The strategy requires a floor.

Chapter 9

Runway Is Freedom in Months

Startups have a word for how long they can survive before they must either earn or die. The word is runway, and it is measured in months. Runway is what you have in the bank divided by what you burn every month. A company with a year of runway can say no to a bad deal. A company with six weeks of runway will sign anything.

You have a runway too. Almost nobody at twenty-four calculates it, and it is the most important number in your financial life, more important than your salary, because it is the number that decides whether you can ever say no.

The formula is the same. What you have, divided by what you must spend each month to stay alive and housed and insured. The answer is a number of months. That number is the exact amount of freedom you own.

The two ways to lengthen it

There are only two ways to make the number bigger. You can make the top bigger, by saving more, or you can make the bottom smaller, by needing less. Everyone tells you about the top. Almost nobody tells you about the bottom, and the bottom is where the power is.

Here is why. Saving is slow and linear. Every month you save, you add one month’s surplus to the top. But cutting a fixed cost is immediate and it compounds. If you cut what you need each month by a quarter, every dollar you already have is suddenly worth a third more months of runway, and so is every dollar you will ever save. You did not earn anything. You simply stopped needing it, and the number moved.

Figure 04

Runway, and the two ways to lengthen it

what you havewhat you must spend each month=months of freedom

Today

Save a quarter more

Need a quarter less

Need half as much

Saving adds to the top once. Needing less multiplies every dollar you already have and every dollar you will ever save.

This is why the strategy says low financial needs are the whole game. Not high savings. Low needs. A person with modest savings and almost no fixed costs has more freedom than a person with a large salary and a life that consumes all of it, and the difference is not comfort. The difference is that the first person can make the offer in Chapter 8 and the second cannot.

Every fixed cost is a month sold

Think of each recurring obligation you take on as a sale of your freedom, priced in months. A car payment is not a car. It is a certain number of months of runway, every year, forever, until the loan is gone. A mortgage is thirty years of months. The insurance on either one is a payment that never stops and never builds anything.

I am not against cars or houses. I am against buying them at twenty-four, before you have a floor, because the obligations you take on then are the exact obligations that will make you unable to treat your job as a customer. A person who must make a payment on the first of the month cannot risk a customer saying no on the twenty-ninth. They stop being a vendor and become a lifeline-holder, and a lifeline-holder will accept almost anything.

The floor

So the first move in Part Three is to find your floor: the smallest monthly amount you could live on, housed, fed, insured against catastrophe, connected and able to work, without being miserable. Not the smallest imaginable. The smallest you could sustain for two years without resentment. That is your floor, and the entire strategy rests on it. The lower it is, the more offers you can make, the more noes you can survive, and the more customers you can walk away from.

Your parents’ floor was high because the world they entered priced the floor low. A house was cheap relative to a salary, insurance was a rounding error, and a job came with a pension. Your world prices the floor differently, and the next chapter is about the three decisions that set it.

Chapter 10

Rent, Ride, Insure the Minimum

Three decisions set the floor for a twenty-four-year-old, and they are the three your parents made almost without thinking, because in their world the decisions were small. In yours they are the whole thing.

Rent

Buy a house, they will tell you. It is the responsible thing to do. Renting is throwing money away.

Here is the first principle underneath that advice, and why it no longer points where it used to. A house is a leveraged asset. You borrow most of its price and you pay for the privilege over decades. In a world where the house reliably rises in value and the costs of holding it are small, the leverage works for you and the advice is right. In a world where the costs of holding it are large and rising, the leverage works against you, because the costs do not care what the house is worth. They arrive every month regardless.

The costs of holding a house are the ones nobody puts in the brochure: the insurance, the taxes, the repairs, the association dues, and on a coast like mine, the second insurance policy for water that the first policy refuses to cover. Every one of those is attached to an asset you had to borrow to buy, and every one of them is a fixed cost that lengthens your burn and shortens your runway.

Renting at twenty-four is not throwing money away. It is paying for exactly the housing you use, for exactly as long as you need it, with no obligation past the lease, and with the ability to move toward the next customer in a month. Mobility is a vendor’s asset. A mortgage is a thirty-year promise to stay put.

Ride

The second decision is the car, and the argument is the same. A financed car is not transportation. It is a payment, plus the full insurance the lender requires, plus fuel, parking, maintenance and the slow conversion of a large asset into a smaller one. Most of those costs run whether you drive or not. The car sits in the lot while you work, costing you runway.

Ride-sharing, transit, a bicycle, your feet, and the occasional rental when you truly need a car for a day all have the property a vendor wants: you pay only when you use them. On a month when you need no car, the cost is zero. No fixed cost on a lifeline-holder’s calendar has that property.

I know this is harder in some places than others. It is easier in a city, easier still in a dense one, and I tell my students that if a job requires a car, that requirement belongs in the price of the job. A car the job needs is a cost of serving that customer, and a good vendor prices the customer’s costs into the customer’s invoice. It does not absorb them.

Insure the minimum

The third decision is the one that has changed most since your parents were your age, and the one almost nobody explains to you plainly.

In their world, insurance rode along. It was a line in the mortgage escrow, a small addition to the car payment, a payroll deduction barely noticed. You decided on the house and the car and the job, and insurance came with them. In your world, insurance is large enough and rising fast enough that it is no longer a rider on the decision. It is the decision. I wrote a whole piece about this for my students, and I will not repeat its figures here, because the figures will be out of date before you finish the book and the principle will not.

The principle is this. Every asset you own that can be destroyed must be insured, and the insurance is a fixed cost that never ends and never builds anything. So the cheapest insurance is to own fewer things that can be destroyed. A renter without a car carries almost no insurance that is not about their own body. That is not recklessness. It is the floor.

For your health, carry real coverage against catastrophe, the kind that keeps one bad year from ending your runway, and carry it in a form that is yours and comes with you when you leave the job. The reason, as Part One explained, is that health insurance was stapled to employment by a run of administrative decisions that had nothing to do with you, and the staple is the thing that keeps most people from ever treating their employer as a customer. If your coverage depends on your job, you cannot leave the job, and a vendor who cannot leave is not a vendor.

The sentence underneath all three

The three decisions share a sentence, and once you see it you will see it everywhere: own the asset, inherit the cost. A house, a car, a boat, a second house. Each comes with a tail of obligations attached to it that run whether you use it or not. The strategy at twenty-four is to own almost nothing that has a tail and to rent the use of everything else, until your floor is low, your log is thick and your first customer is paying you on terms you set. Then, if you still want the house, buy it with a vendor’s money instead of a lifeline-holder’s.

Figure 05

Own the asset, inherit the cost

House

• Mortgage
• Insurance, and a second policy for water
• Taxes and dues
• Repairs

Rent instead: pay for the housing you use, for as long as you use it.

Car

• Payment
• Full coverage the lender requires
• Fuel, parking, maintenance
• Sits in the lot while you work

Ride instead: pay only on the days you need it.

The job’s plan

• Coverage that ends when the job ends
• A match that vests on a schedule built to keep you

Carry your own: coverage against catastrophe that comes with you.

Chapter 11

The Floor Is Your Price

Now the floor does its real job, which is not saving you money. Its real job is to let you say no.

The ability to refuse is the only thing that separates a vendor from an employee. An employee can refuse a task and be fired. A vendor can refuse a customer and still eat. Everything in Part Three has been aimed at this one capability, because every other part of the strategy depends on it. You cannot make the offer in Chapter 8 unless you can afford to have it refused. You cannot price your work properly unless you can walk away from a bad price. You cannot keep your own AI, your own log and your own method unless you can lose the customer who demands you give them up.

So the floor is not a budget. It is your negotiating position, and it is the strongest one there is, because it is the one position no customer can take away from you.

Price from the floor up

Here is how an employee thinks about pay: what is the salary for this role, and how much more can I get? The anchor is the market rate for a job title, which is a number set by other people about other people.

Here is how a vendor thinks about price: what does it cost me to deliver this, what is it worth to the customer, and what is the lowest price at which I would rather do it than not? The third question is the floor speaking. If your monthly need is small, the lowest price at which you would rather do the work than not is low, which sounds like weakness and is the opposite. It means you can afford to say yes to the customers you want and no to the rest, and a vendor who can say no gets paid more than a vendor who cannot, because the customer can feel the difference.

In Chapter 8 you wrote down what it costs the company to employ you beyond your salary. That number matters here. When you make the offer, the company is not comparing your retainer to your salary. It is comparing it to your salary plus everything else it spends to own your Tuesdays. A retainer that looks expensive next to a salary often looks cheap next to the whole cost of an employee, and you should price it there, from the customer’s side, not from your own need.

The hour is a cost. The finished thing is the product.

You will be asked, constantly, what you charge an hour. Answer the question; people need a number. But never confuse the unit on the invoice with the thing being sold. The hour is how you measure your cost. The finished thing is what the customer is buying. A roofer may quote by the day, but the customer is buying a roof, and if the roof leaks the day rate does not matter.

So price in whatever unit the customer understands, and deliver in outcomes. Over time, as your log grows and your customers trust it, you will find you can price the outcome directly, which is where every mature vendor ends up. That takes years. In the first year, keep it simple and keep it honest. Charge enough that the floor is covered with room to spare, and then earn the right to charge more by delivering.

The first no

At some point a customer, perhaps your first employer, will ask you for something that crosses a line. To hand over your AI and use theirs. To stop keeping your log. To make yourself available in a way that turns the outcome purchase back into a time purchase. To accept a price that does not cover the floor.

The first time you say no is the moment the strategy becomes real. Up to then it has been a posture and a spreadsheet. The first no proves the floor exists. Say it kindly, say it plainly, and offer the alternative you can live with. Most customers, including most employers, respect a clear no far more than they respect a resentful yes, because a clear no tells them you will also say a clear yes, and that the yes will mean something.

If they walk, your floor holds you. That is what it is for.

Chapter 12

Manager, Not User

A vendor shows up with tools. A roofer brings ladders; a photographer brings cameras; an accountant brings software and a method. The tools are part of what the customer is buying, and the vendor would no more leave them at the door than a surgeon would borrow the hospital’s hands.

Your tools are your AI and everything you have built around it. This part of the book is about bringing them, and it starts with a correction to the word everyone uses.

You are not AI’s user. A user is a person in a relationship with an object: you use a hammer, a spreadsheet, a car. AI is not an object in that sense. It is an entity that can take an outcome you describe, figure out a path, use tools along the way, and come back with finished work. You do not use something like that. You manage it. You delegate to it. The right word for you is manager, and the right word for it, I have come to think, is delegate: not an agent, named for what it does, but a delegate, named for who sent it, bound by a mandate, and owing you a report when it comes back.

This is not a quarrel about vocabulary. It changes how you work in three ways that matter to the strategy.

Describe the state change

A user gives instructions: open this, search for that, make a table with these columns. A manager describes an outcome: what will be true when this is finished that is not true now, and how we will both know. My two favorite prompts are five words long between them, and neither tells the AI how to do anything. You do it. You decide. The first sends the work back when it returns as a to-do list addressed to you. The second refuses to answer questions a competent delegate could answer alone.

Notice that this is exactly the posture you are taking toward your employer in reverse. You want to be given outcomes, not steps, and to return finished work. Your AI wants the same from you. A person who manages their AI well is practicing, every day, the relationship they are asking their first customer to have with them.

Send the small matters down

A manager does not do the small matters. A manager sends them down and keeps only what is worthy of their attention. The test is not difficulty and it is not time. It is respect: is this matter worthy of a human being, and is this person worthy of mine? If the matter is a form, a search, a reconciliation, a first draft or a schedule, it goes down. If the matter is a decision only you can make, a risk the customer must know about, money, or anything that goes out under a name to another person, it stays with you.

And one rule above all the others: never send your AI to deal with a person who is worthy of your attention. Your customer is such a person. Your AI can prepare you, brief you, clear your morning and write the first version of everything. When the customer is in the room, you are in the room.

Trust is a record

The third change is the one students find hardest. You will want to check everything your delegate does, the way a new manager hovers over a new hire. Do that at first. Then stop, deliberately, for the categories where the record shows it is reliable, and keep checking only where it is not. Trust is not a feeling you have about your AI. It is a record of what it has delivered correctly, kept the same way you keep your own log. Where the record is long and clean, you delegate without looking. Where it is short, you look. The frontier of what it can do moves every few months, so keep a list of what it could not yet do, with dates, and try again. The list expires.

Chapter 13

Bring Your Own AI

On the first morning of the job, the company will hand you a laptop with its own AI already on it. It will be capable and polite, and it will know nothing about you.

I have argued for a long time that personal AI will beat enterprise AI, and that every business should have a policy that says, in effect, bring your own. The argument for the business is in another piece of mine. The argument for you is simpler, and it is the reason this chapter exists.

The model of you is the asset

An AI that works beside you for a year learns things no company could or should know. When you are sharp and when you are not. Which decisions surprise you and which you settle in your sleep. How you like work to come back. What you have already tried. Where your own records are. That knowledge is built mostly outside the job, from your whole life, and it compounds every week it keeps watching you work.

Now imagine leaving it behind. The company’s AI, the one on the laptop, has learned all of that about you over two years, and the day you leave, it stays. Your next first morning begins at zero, not because you forgot anything, but because the thing that knew you did not come with you. You walked out with your experience and left behind the model of it.

A vendor would never accept that. A vendor’s tools go with the vendor. Your AI is a tool, and the model it holds of you is the part of the tool that took years to make. It must be yours, on an account you own, and it must travel.

The company’s half and your half

This does not mean refusing the company’s systems. The business owns the situation: the deadlines, the systems of record, the permissions into them, the work product. Your AI should connect to those through permissions the company grants and can revoke. What you own is the other half: the model of you, your method, your record of decisions, your log. The work stays with the business. The model of you leaves with you. A good policy says that in one sentence, and the best companies will soon put the sentence in the offer letter.

Figure 06

The seam: the company’s half and yours

The company brings

• The situation: deadlines, windows, service promises
• The systems of record and the permissions into them
• The work product, which stays
• A default AI for anyone who arrives without one

the sort happens here

You bring

• An AI that knows you, on an account you own
• Your method, refined over jobs
• Your delivery log
• The judgment for the matters that reach you

The work stays. The model of you leaves. Put that sentence in the offer letter.

Until they do, you can live the arrangement yourself. Do the company’s work in the company’s systems. Keep your AI, your log and your method on your own account. Connect the two with the narrowest permissions that let the work get done. When you leave, revoke and go.

The employer is a tenant

If you have read Super Admin, you know the frame: you are the one who draws the table. The company is a tenant in your stack, with its own scoped space, its own permissions, and a clean exit. Everything you build for the customer lands in the customer’s space and belongs to the customer. Everything you have learned about how to do the work lands in yours. The seam between them is the thing you are protecting, and it is also the thing that makes the whole arrangement safe for the company, because they know in advance exactly what leaves with you and what does not.

A student once asked me whether this was disloyal. It is the opposite. The disloyal arrangement is the one where the company quietly keeps the model of a person it did not pay to build, and the person leaves poorer than they arrived. Drawing the line clearly is what makes it possible to be generous on both sides of it.

Chapter 14

The Delivery Log Is the Resume

For about a century, you have been asked to make yourself readable to employers in a format they invented for their own convenience: the resume. Title, company, dates, bullet points. The format exists because an institution can act only on what it can read, and for most of that century the cheapest way to read a person was to make the person pre-shape themselves into a page.

That is ending, and the delivery log you started in Chapter 7 is what replaces it.

Why the resume was a format

A resume does not say what you delivered. It says where you sat and for how long, and it lets you describe, in your own flattering words, what you were responsible for. Everyone knows this. Hiring managers read resumes the way they read advertising, discounting as they go, and then try to find out the one thing the resume cannot tell them: what actually happened when this person was handed work.

The log tells them. Date, delivery, who accepted it, what changed. It is not an argument about you. It is a record, and a record can be checked. A first customer who will say “yes, that was delivered, and it was right” is worth more than any title, because the title describes a seat and the customer describes an outcome.

Figure 07

A resume line and a log line

Resume

Operations Coordinator, 2027-2028

Responsible for scheduling, vendor management and process improvement across a large team.

A seat, a span of time, and adjectives.

Delivery log

2027-03-04 Posted April schedule, all shifts covered, two moves. Accepted by Dana. Overtime stopped.

2027-03-19 Vendor audit shipped: overcharge found in three months, refund requested. Accepted by Dana. Refund received 04-02.

Dates, deliveries, acceptance, what changed. Checkable.

AI reads the log raw

Here is what has changed. For the first time, the reader on the other side does not need you pre-shaped. An AI can read a three-year log of finished deliveries, in whatever order you kept it, and tell a hiring manager or a prospective customer exactly what you have shown you can do, how often, and how it was received. It can read your customer’s testimony next to it. It does not need the bullet points. It needs the record.

So the burden of being readable is changing hands, and the person who kept a log is the person the new reader can read best. The person who kept only a resume has given the new reader nothing to work with but adjectives.

Keep it like a vendor

A few rules for the log, from someone who publishes one.

Record deliveries, not activity, and record who accepted each one and what they said, because a line without acceptance is a claim and a line with acceptance is evidence. “Attended” is not a line. “Delivered, accepted by Dana” is.

Record what changed. The schedule got posted and the overtime stopped. The report shipped and the vendor refunded the overcharge. The state change is the thing that was bought, so the state change is what the line should say.

Keep it in a file you own, in plain text, on your own account. Your AI should write most of the lines for you, from the work it watched you deliver; your job is to make sure the acceptance is real. And keep it honest. The log is valuable only because it can be checked. A padded log is a resume with extra steps.

My own business keeps a log of every piece of work it has delivered and publishes it, one delivery at a time, as the public record of what we do. When someone asks what we can do for them, we do not describe it. We send them to the log. I want the same for you. When your first customer asks what you can do, you should not have to say. You should be able to show.

Chapter 15

Read the Contract

Everything in this book so far has been about posture, money and tools. This chapter is about paper, because the paper you sign on the first day can quietly cancel the strategy, and almost nobody at twenty-four reads it.

I am not a lawyer, and this is not legal advice. The rules differ by state and they change every year. What I can do is tell you which three clauses to find, what each one is for, and what question to ask about it, so that when you do talk to a lawyer, or to the company, you are asking about the right things.

The three clauses

The first is the invention assignment, sometimes called a proprietary information and inventions agreement. It says that what you create in the course of your work belongs to the company. Read it for one thing: does it reach work you do on your own time, with your own tools, unrelated to the business? Some agreements are written broadly enough to try. About ten states, California the best known among them, have laws that limit how far such a clause can reach into an employee’s own time and own equipment; most states do not. If your log, your method and the things your AI has built for you could be read as company property, the strategy has a hole in it. Ask the question before you sign.

The second is the outside work clause, sometimes called a conflict of interest or moonlighting policy. These are common and, in their ordinary form, reasonable: tell us before you take outside work, do not work for a competitor, do not use our confidential information, do not let it interfere with your job. The practical point is disclosure. A vendor posture does not require hiding anything, and the day you take a second customer, the company should hear it from you first, in writing, in the form the policy asks for. A flat ban on any outside work is a different thing, and a weaker one in many states, but if it is in your agreement you need to know it is there.

The third is the restrictive covenant: the noncompete, the non-solicit, the non-disclosure. The noncompete is the one that matters most to the conversation in Chapter 8, because it can forbid you from doing for a customer, as a vendor, the exact work you did for them as an employee. Here the ground is moving. A federal attempt to ban noncompetes nationwide never took effect; it was set aside by a court before its start date and later withdrawn, and as I write, the question is handled state by state, with the federal regulator pursuing individual employers one case at a time. A handful of states ban noncompetes outright. My own state, South Carolina, disfavors them but, as I understand it, will enforce one that is reasonable in time and territory, protects a legitimate business interest, and meets a few other tests. Find out which kind of state you are in. Then read the clause, and the non-solicit and confidentiality clauses beside it, for whether they would cover your first employer becoming your first customer. Whether they do depends entirely on how they are written. That is the third question below, and it is the one worth an hour of a lawyer’s time.

Three questions, asked plainly

You do not have to negotiate any of this on the first day, and I would not. The agreement is usually a standard form and the person handing it to you did not write it. But you can ask three questions, in writing, and keep the answers.

Does the invention assignment reach work I do on my own time, with my own equipment, unrelated to the business?

If I take outside work later, what is the process for disclosing it, and who approves?

Does the noncompete, if there is one, cover my continuing to serve this company as an independent vendor after I leave?

The answers tell you how much of the strategy is open to you here, and asking them tells the company something true about you: that you read what you sign.

When the day comes

If you do convert from employee to vendor, there is a second set of paper. The company will have to satisfy itself that you are, in the eyes of the tax authorities and the labor laws, independent, and either of you can ask the tax authority to rule on it. Most of the tests ask about control: who decides how the work is done, who provides the tools, whether you can take other customers, whether the relationship has an end. A vendor who sets their own hours, brings their own AI, serves more than one customer and delivers outcomes rather than attendance is answering those questions the right way. But a few states apply a stricter test, under which doing the company’s ordinary work makes you an employee no matter how independent you are, and the federal rule has been rewritten more than once in the last few years. This is the one place in the book where I will tell you to pay for an hour of a lawyer’s time before the day comes.

If the arrangement passes, you will pay your own taxes, including the share your employer used to pay on your behalf, and you should price that in. You will receive a tax form instead of a pay stub. None of this is hard. It is unfamiliar, and most of it is paperwork your AI can prepare for you to check.

What matters is that you built the relationship, from the first week, so that it would pass these tests in substance. The paper follows the practice. It cannot substitute for it.

Chapter 16

The Boundary Moves

In 1998, two researchers at MIT published an article in the Harvard Business Review with a title that sounded like science fiction and an argument that was pure Coase. It was called The Dawn of the E-Lance Economy, and it opened with a young Finnish programmer who had organized strangers, over a network, to build an operating system without a company. The authors, Thomas Malone and Robert Laubacher, drew the conclusion that Coase’s theory required. If networks were making it cheap to find people, describe work and coordinate it, then the cost of using the market was falling, and the boundary of the firm would move inward. Large permanent companies would give way to shifting networks of independent people, forming around a piece of work and dissolving when it was done.

They were right about the direction. They were early about the timing, by about twenty-five years, and the reason they were early is the most important thing in this part of the book.

Why it waited

In Chapter 2 I named three costs of using the market: finding, describing and checking. Networks lowered the first nearly to zero. The second fell partly; it became easy to post a description and hard, as ever, to write a good one. But the third cost did not move, and it is the one that actually keeps work inside companies: the cost of knowing whether the work is any good.

When you hire an employee, you do not have to check each piece of work before you pay for it. You pay for the time and you supervise. When you buy a piece of work from a stranger on a network, you have to specify it exactly, and then you have to check that what came back is what you asked for, and checking is expensive, because checking well often takes nearly as much skill as doing. For twenty-five years, that cost kept most serious work inside the building. The e-lance economy arrived, but it arrived for the kind of work that was easy to check: a logo, a translation, a few hours of code with tests attached. The rest stayed with employees.

What has changed is that finished work has become both cheap to produce and cheap to check. A person with a capable AI can deliver a complete, tested outcome, not a draft, and the person buying it can have their own AI verify it against the stated outcome in minutes. The specification is the test. When the work comes back, you do not read it the way you read an employee’s draft, hunting for what went wrong. You check it against the state change you asked for. Done or not yet.

That is the third cost falling, and it is why the conversation in Chapter 8 is becoming possible for ordinary people doing ordinary work, rather than only for the few whose work was always easy to check.

The shamrock

Nine years before Malone and Laubacher, an Irish management writer named Charles Handy had drawn the shape of the firm that results. He called it the shamrock organization, in a 1989 book called The Age of Unreason, and it has three leaves. The first is the professional core: the people who carry the organization’s essential knowledge and are paid, well, to stay. The second is the contractual fringe: outside people and firms, often former employees, who do work the core does not need to do itself and are paid for what they deliver. The third is the flexible labor force: part-time and temporary workers who absorb the peaks.

Handy’s second leaf is the strategy in this book, drawn from the company’s side. Often former employees. Paid for what they deliver. That is what you become when your first employer becomes your first customer, and Handy saw, in 1989, that companies would want it as much as the people did. He also saw what it would mean for the person: a working life built as a portfolio of customers and pieces of work rather than a single ladder in a single building.

Figure 08

The boundary moves inward

The firm, 1956

The firm, 2026

The chargedecisions that surprise, people worthy of attention

• Outside the core: finished work, bought from vendors who bring their own AI.
• Inside: whoever carries the charge. That is what a company is for.

What the firm keeps

If the boundary moves inward, what stays inside? Coase gives the answer in principle: whatever is still cheaper to make than to buy. I think, in practice, that comes down to one thing.

A company keeps the people who carry the charge. I have written elsewhere about what I mean by that: the decisions that carry real surprise for the person making them, the ones that cannot be settled by rule, and above all the people, customers, partners, colleagues, who are worthy of a human being’s own attention and would be dishonored by a machine. Finished work can be bought. Judgment on behalf of the firm, exercised by a person the customer knows, cannot, or not yet, and the companies that understand this will shrink to it deliberately, keeping the few who carry the charge and buying everything else from vendors who bring their own AI.

That is the market you are walking into. It is not a market that has stopped hiring. It is a market where the first job is the beginning of a customer relationship, whether or not either side has noticed yet, and where the person who notices first, on the first morning, has an advantage that compounds for the rest of their working life.

Where I could be wrong

I will give the objection its due here and again in the ledger, because it is a strong one. The technology that makes it cheap to buy finished work from outside also makes it cheap to manage work inside. If an AI can specify, assign and check the work of a hundred vendors, it can do the same for a hundred employees, and the cost of organizing inside the firm falls along with the cost of using the market. Coase’s theory does not say which cost falls faster. It only says the boundary will move toward whichever one does.

I am betting on the market, for the reasons in Part One: the tools are in the worker’s hands, the method travels, and people will not let an AI see their whole life unless it answers to them. But the strategy does not need me to be right about that. A person with a low floor, a thick log and their own stack can work on either side of the boundary. That is the point of building all three.

Chapter 17

Five Predictions, with a Ledger

I end every forecast the same way, with a ledger. For each prediction I state what is already true, what has to happen for it to come true, and where I am probably wrong. The last column is the one I care about most. A prediction without a failure condition is not a prediction; it is a mood.

Epilogue

Two Offer Letters