Trillions: The Consumer Economy After Peak Human

Prologue

The Receipt

The line items are small and strange. A connector to a calendar. Rows in a database. A few thousand calls to a mapping service. A month of storage. A transcription service that ran for eleven minutes on a Tuesday. Tokens, in batches, all month long. Each one was a decision, and each decision was made by something that works for me.

My AI agents spend more than I do now. I want to say that plainly, because it sounds like a joke and it is not. They buy connectors, databases, API usage, services and tokens. They buy them without asking me, inside limits I set, for work I gave them. When I look at the month, my own purchases are a short list: groceries, gas, a dinner out, a gift. Theirs is the long one.

I am one man with a modest stack. I am early, not unusual. In a few years this will be the ordinary shape of a household’s month, and of a business’s, and then of a city’s.

Here is the thought that started this book. For thirty years we have been told to fear the peak. The human population will stop growing, sometime this century, and then it will fall. Fewer people means fewer consumers. Fewer consumers means less demand. Less demand means the long, gray stagnation that Japan has been living in for a generation, coming for everyone.

That fear rests on one assumption: that a consumer is a person.

It never was. And by 2056, the year I will use as the horizon in this book, the number of consumers in the economy will not be counted in billions. It will be counted in trillions.

This book is about what those trillions will buy. Fashion? Land? Subscriptions? Toothpaste? Batteries? I will answer each of those questions directly. But first I need to show you why the count is about to leave the head count behind, because once you see the principle, most of the answers fall out of it on their own.

Chapter 1

The Wrong Census

Every forecast of demand starts with a census. How many people, of what age, where, with what income. Multiply, and you have a market. It is such a natural move that nobody notices it is a move at all.

The demographers have been quietly revising that census downward for years. Families are smaller nearly everywhere. Some of the largest countries in the world have already passed their peak and begun to shrink. The United Nations expects the global count to top out in the 2080s. Other demographers think it comes sooner, in the late 2050s or early 2060s. I use 2056 as my horizon in this book. The exact year does not matter to the argument. What matters is that there will be a year, and after it the head count goes down.

The fear that follows is simple. An economy built on growing demand meets a shrinking number of people to demand things. Shops close because there is nobody to buy. Houses empty. Pension systems strain because fewer workers support more retirees. Japan is the usual example: an aging, shrinking country whose economy has spent a generation struggling to grow.

I do not dismiss any of that. The pension problem is real. The aging problem is real. But the demand problem, the fear that there will be no one left to buy, rests on a definition that was wrong before AI and is about to become wrong in a way nobody can ignore.

What a consumer actually is

Economists have a precise phrase for the kind of wanting that moves markets. They call it effective demand: a want backed by the ability to pay. Wanting a boat is not demand. Wanting a boat and having the money to buy one is.

I want to add one more piece, because it is about to matter more than the other two. Effective demand also needs authority. Someone has to be allowed to decide. A child may want the boat and may even have the money in a savings account, but the child does not have the authority to buy it. The parent does.

So a consumer, stated from first principles, is three things together:

• A want: something it needs or is trying to accomplish.
• A wallet: money it can spend, its own or someone else’s.
• Authority: permission to turn the want into a purchase without asking.

Notice what is not on that list. A body. A face. A pulse. A vote.

We have always had non-human consumers

Once you define a consumer this way, you see something that was true all along. The economy has been full of non-human consumers for centuries. We just called them by other names.

A company is a consumer. It wants things, it has a budget, and it has people with the authority to spend that budget. The purchasing department of a hospital buys more gloves in a week than you will buy in a lifetime. A city buys asphalt, salt, uniforms and software. A government buys ships. Economists call most of this intermediate consumption, or business-to-business trade, and in most economies it is larger than everything households buy in shops put together.

None of these buyers is a person. Each is a want, a wallet and an authority, organized into an entity. When we counted consumers by counting heads, we were already undercounting. We were counting the people at the end of the chain and ignoring the much larger tangle of buyers in the middle.

The demand problem rests on a definition that was wrong before AI and is about to become wrong in a way nobody can ignore.

The new entity

What is new is not that non-human things buy. What is new is the size of the thing that can buy.

A company is a big, expensive way to bundle a want, a wallet and an authority. It takes lawyers, accountants, offices, a bank account and a payroll. Because it is expensive, there are relatively few companies. You can count them.

An AI delegate is a small, cheap way to bundle the same three things. It has a want: the task it was given. It has a wallet: a budget delegated to it. It has authority: a mandate that says what it may buy without asking. It costs almost nothing to create, and you can create as many as you have tasks.

That is the whole shift, stated in one paragraph. The unit that can consume is shrinking from the organization to the task. And there are vastly more tasks than there are organizations, or people.

Chapter 2

Ford’s Turn, and the Next One

There is an old story about Henry Ford that I like because it is about this exact question: who counts as a consumer?

In 1914 Ford raised wages at his plant to five dollars a day, roughly double the going rate. A dozen years later he cut the work week to five days. He said a number of things about why, and historians still argue about how much was generosity, how much was turnover, and how much was strategy. But the effect is not in dispute. Workers who had been paid too little to buy what they made, and who had too few hours off to use it, became people who could buy a car and had a Saturday to drive it.

I wrote about this recently because I think it is one of the most important turns in economic history. Ford did not invent a new product. He invented a new consumer. He took a group of people who were inputs to production and gave them the wallet and the time to become buyers of its output.

The history of consumption is a history of new consumers

Look at the last two centuries this way and a pattern appears. Every great expansion of demand came from handing purchasing authority to a new kind of unit.

• The household gave way to the wage earner, who could buy for himself.
• The wage earner was joined by the wage-earning woman, whose spending was her own.
• Teenagers got allowances and part-time jobs, and an entire industry of music, clothes and cars grew up to sell to them.
• Credit cards gave nearly every adult a line of authority to spend money they did not have yet.
• The internet put a store in every pocket and let anyone buy anything at three in the morning.

Each of these is the same move. A unit that could want, but could not pay or could not decide, was given the missing piece. Demand grew not because people suddenly wanted more but because more things were allowed to act on what they wanted.

The next turn

The next turn is the largest one yet, and it is already underway on my statement every month.

We are handing the wallet and the authority to things that are not people at all. Not companies, which are slow and expensive. Delegates: small, fast, cheap entities that each carry one want and a slice of a budget. Ford made the worker a consumer. AI is making the task a consumer.

There is a difference worth naming. Ford’s new consumers were people who wanted things for themselves. A delegate does not want anything for itself, at least not at first. It wants what its task requires. But as you will see in Part Two, a task requires a surprising amount, and much of what it requires is consumed by the delegate itself, used up in the doing.

Ford made the worker a consumer. AI is making the task a consumer.

Chapter 3

The Buyer Shrinks to the Size of the Transaction Cost

Why trillions? Why not billions, one delegate per person, or a few each? The answer comes from an essay written in 1937 by a young economist named Ronald Coase, and it is the most important principle in this book.

Coase asked a question so basic that nobody had thought to ask it. If markets are so efficient, why do firms exist at all? Why does anyone hire employees and tell them what to do, instead of buying every piece of work on the open market, task by task?

His answer was transaction costs. I built my last book, First Customer, on the same essay, because it explains the job. Here it explains the buyer. Every purchase costs something beyond its price. You have to find a seller, compare offers, negotiate, write the terms, check the work, pay, and handle the dispute if something goes wrong. For a large, rare purchase, those costs are small relative to the price. For a small, frequent one, they swamp it. So instead of buying each small piece of work separately, we bundle the pieces inside a firm and manage them by command. The firm grows until the cost of managing one more thing inside it equals the cost of buying it outside.

Run Coase backwards

Here is the move that matters. Coase explained why buying gets bundled when transactions are expensive. Run his logic the other way. When transactions become nearly free, buying unbundles.

That is exactly what AI delegates do to transaction costs. A delegate can find a seller in a second, read every offer, compare them against the task, accept the terms, pay a fraction of a cent, and check the result, all without a human spending a minute of attention. The search cost, the negotiation cost and the monitoring cost all collapse at once.

When that happens, the buyer shrinks. It shrinks from the firm, which bundled thousands of purchases under one roof, to the household, to the person, and then below the person, to the individual task. The task becomes the natural unit of buying, because there is no longer any reason to bundle it with anything else.

The buyer shrinks to the size of the transaction cost. When the transaction cost approaches zero, the buyer approaches the size of a single task.

The door the web left open

There is a small, lovely piece of history here. When the engineers who built the web wrote down the codes a server can send back, they reserved one for payment. Code 402: Payment Required. In the specification of 1997 it was marked for future use, and for almost thirty years nobody used it. A person cannot be asked to pay a tenth of a cent every time a page loads. The attention would cost more than the money.

That code is being used now, by machines, for exactly what it was reserved for. A delegate asks for a piece of data. The server answers: payment required, and here is the price. The delegate pays and gets the data, in less time than it took you to read this sentence. The web had a door for tiny purchases all along. It was waiting for a customer small enough to walk through it.

The arithmetic of trillions

I said at the start that I would not lean on statistics, and I won’t. But the arithmetic of trillions is not a statistic. It is multiplication, and you can do it yourself.

Take the people. Call it nine or ten billion around the peak. Now ask how many standing delegates a person will have. If I count every scheduled task, every connected service that runs on my behalf and every small agent watching something for me, I already have more delegates than I can list from memory. My students, who are about twenty-four, will have more than I do, because they will never have known a world without them. A hundred standing delegates per person is not a fantasy. It is a modest stack ten years from now.

Ten billion people with a hundred delegates each is a trillion consumers. And I have not counted any businesses, which will run far more delegates than any household. I have not counted cities, governments, hospitals, schools. I have not counted the machines that move: the cars, robots, drones and appliances that buy their own charging, parts and maintenance. And I have not counted the short-lived delegates, the ones spun up for a single job and retired an hour later, which will outnumber the standing ones the way mayflies outnumber turtles.

You can argue with any number in that paragraph and the conclusion survives. Make it ten delegates per person instead of a hundred and the businesses alone will carry you past a trillion. The count of consumers in 2056 is not going to be measured in billions, and it is not going to fall when the head count falls. It is going to keep climbing, because it is not tied to births. It is tied to tasks, and tasks multiply every time something gets cheaper to do.

Figure · Chapter 3

The arithmetic of trillions

10 billionpeople near the peak

×

100standing delegates each

=

1 trillionconsumers

• plus every business’s delegates
• plus cities, hospitals, schools and governments
• plus machines that buy their own charging, parts and repairs
• plus short-lived delegates made for one job and retired an hour later

But is it real demand?

Here is the strongest objection, and I want to face it now rather than at the end. Every dollar a delegate spends is ultimately someone’s dollar. My delegates spend my money. A company’s delegates spend the company’s money. So isn’t this just a recount? The same human money, routed through more hands?

Partly, yes. And the recount alone matters, because who decides changes what gets bought. When a hospital moved its purchasing from individual doctors to a central department, the money was the same, but the gloves were different. When delegates take over a purchase, the same money buys different things, from different sellers, for different reasons.

But it is more than a recount, for a reason that is easy to miss. Look again at the list on my statement. Tokens. Storage. Database rows. Calls to a mapping service. Those things were not bought for me to use. They were used up by the delegates in the course of their work. The delegate consumed them, in the oldest sense of the word: it used them up. A delegate is not just a buyer that passes goods along to a human. It is a consumer in its own right, with its own needs, which have to be met before it can do anything at all.

That is what Part Two is about.

Chapter 4

Hunger

In 1857 a German statistician named Ernst Engel studied the budgets of working families in Belgium and found a pattern so reliable it became a law. The poorer the family, the larger the share of its money that went to food. As income rose, spending on food rose too, but more slowly than income, so its share fell. The rich spend more on food than the poor, and a smaller part of their money.

Engel’s law is really a statement about hierarchy. Every consumer has a subsistence basket, the things it must buy just to keep going. Only after that basket is filled does money flow to anything else. A family eats before it buys a radio.

So the first question to ask about a new kind of consumer is: what does it eat?

The delegate’s subsistence basket

A delegate eats four things.

• Compute: the thinking itself, bought as tokens or as time on a machine.
• Energy: the electricity under the compute, which every token ultimately burns.
• Memory: storage for what it knows, what it did, and what it must remember next time.
• Connection: the network that carries its requests and the data that comes back.

Everything a delegate does consumes all four. It cannot think without compute, cannot compute without power, cannot continue a task without memory, and cannot reach anything without connection. This is its food, water, shelter and air. It is the line on my statement that never stops: tokens, in batches, all month long.

There is a difference from a human family that matters a great deal. A person eats about the same amount whether they are idle or busy. A delegate eats in proportion to its work. Give it more to do and it eats more. Give it a harder problem and it eats more again. Its appetite is not fixed by biology. It is set by ambition.

Cheaper food, more eating

The usual expectation is that as compute gets cheaper, spending on it will fall. That expectation is wrong, and an English economist named William Stanley Jevons explained why in 1865.

Jevons was studying coal. The steam engines of his day were becoming far more efficient, burning much less coal for the same work. People assumed Britain would therefore use less coal. Jevons argued the opposite. When something becomes cheaper to use, people find many more uses for it, and total consumption rises. Efficient engines made steam power worth using in places it had never been worth using before. Coal consumption soared.

Compute follows Jevons exactly. Every time the price of a unit of thinking falls, a whole new class of tasks becomes worth delegating. A task that was not worth a dollar of thinking is worth a tenth of a cent. A task not worth a tenth of a cent is worth a hundredth. The price falls, the number of delegates rises, and the total eaten goes up.

A person’s appetite is fixed by biology. A delegate’s appetite is set by ambition.

Hunger is the foundation

This is why I say the delegate is a consumer in its own right and not merely a courier for human demand. Before it buys anything for you, it must feed itself. And because its hunger scales with work and work expands as prices fall, the subsistence basket of trillions of delegates becomes one of the largest markets in the world: power plants, chips, data centers, cables, cooling. It is already the reason the largest companies on earth are spending more on construction than at any time in their history.

We will come back to this when we get to batteries and land. For now, hold onto the shape. Engel’s first rung for the delegate is energy and thought. It will spend more on that rung every year, not less.

Chapter 5

Tools and a Place to Live

Once a family has eaten, it spends on shelter and tools: a place to live, things to work with. The delegate’s second rung looks the same.

A delegate with only compute is like a brilliant person locked in an empty room. It can think, but it cannot do anything. To act in the world it needs tools, and tools are bought.

• Connectors to the places it must work: a calendar, an inbox, a ledger, a store.
• Databases to keep its records in a form it can search and trust.
• Services that do one thing well: mapping, transcription, translation, verification, search.
• Data it does not have: a price list, a weather feed, a filing, a map, a review history.
• Identity: proof of who it works for and what it is allowed to do.

That is my statement again, almost line for line. The connector to the calendar. The rows in the database. The calls to the mapping service. The transcription that ran for eleven minutes. These are the delegate’s hammer and saw, its workbench and filing cabinet.

The stack is its home

I have written before that the right way to think about personal AI is not as an app but as a stack: the layers of memory, skills, connections and permissions that make an AI yours. The stack is where a delegate lives. It is its house.

And like any house, it has running costs. Rent on storage. Fees for the connectors. Subscriptions to services. Every one of these is a recurring purchase made, increasingly, by the delegates themselves, because they are the ones who know what they need. When my delegate decides a task needs a better transcription service, it is in a better position to choose than I am. It knows the task, it has tried the alternatives, and it can compare them in a second.

Why delegates prefer the meter

Here is something you will see again in Chapter 11. When humans buy tools, they prefer subscriptions. A flat monthly fee is easy to think about. You do not have to decide every time you use the thing. The subscription is a way of saving human attention, which is the scarcest thing a person has.

A delegate has no such scarcity. It can make a decision every millisecond without fatigue. So it does not need the subscription’s protection from deciding. It would rather pay for exactly what it uses, call by call, because paying for unused capacity is waste, and it can see the waste in a way we cannot.

This is why the tiny-payment door the web reserved in 1997 is finally being used. The meter was always the more honest way to price a tool. We just could not stand to be metered. Delegates can.

The subscription was a way of saving human attention. A delegate has attention to spare, so it would rather pay by the meter.

Chapter 6

The Luxury Good Is You

Engel’s law has a top as well as a bottom. Once subsistence is met and the tools are bought, money flows toward what is scarce, special, and not strictly necessary. For people, that is the realm of luxury. So what is luxury for a delegate? What does a machine buy when it has met its needs and wants the best?

I have thought about this for a long time, and I believe the answer is: a human being.

The scarcest input

Everything on the delegate’s first two rungs is getting cheaper. Compute falls in price. Storage falls. Data falls. Services multiply and compete. What does not get cheaper, what becomes more scarce every year as the delegates multiply, is human attention.

There are things a delegate cannot do for itself, not because it lacks intelligence but because of what the thing is.

• Judgment that someone is accountable for: a decision a person will stand behind.
• Witness: a human who saw it happen and will say so.
• Signature: a commitment that binds a real person to a real promise.
• A face: the trusted human presence that makes a stranger willing to believe.
• Taste: a choice that matters precisely because a particular person made it.

When a delegate needs one of these, it cannot manufacture it. It has to buy it from a human. And as the number of delegates rises into the trillions while the number of humans levels off and falls, the price of human attention, bought by machines, has only one direction to go.

The inversion

For the whole of the internet era, the business model was to capture human attention and sell it to advertisers. I wrote a book about where that goes, The Zero-Attention Economy, because I believe the attention economy is ending. When delegates do the browsing, the comparing and the buying, there is no human eye on the page to capture. The ad has no one to look at it.

But attention does not disappear. It moves up the hierarchy. It stops being the thing that is harvested cheaply from people scrolling and becomes the thing that is purchased dearly by machines that need a human to judge, witness, sign or show up. The attention economy sold your attention to companies. The economy that replaces it will buy your attention on behalf of machines, and pay you directly.

The attention economy sold your attention to companies. The economy that replaces it will buy your attention on behalf of machines, and pay you directly.

Be the face

This is why I tell my students to be the face that their customers trust. Every digital product and service is going to be made, sold and bought by machines with nothing to lose. Somewhere in each chain there will need to be a person who carries trust, liability and responsibility. That person is the luxury good. That person is what the delegate cannot make and must buy.

It is also, I think, the right answer to the fear that AI will leave people with nothing to do. In an economy of trillions of non-human consumers, the human being is not obsolete. The human being is the top shelf.

Chapter 7

Toothpaste

Will delegates buy toothpaste? Not for themselves. A delegate has no teeth. But they will buy nearly all of it, for us, and that will change toothpaste more than anything since the tube.

The bound customer

The analysts who first wrote about machine customers had a name for the earliest kind. They called them bound customers: machines that buy one thing, by a rule a person set. The printer that orders its own ink when the cartridge runs low is the classic case. It is not very smart. It does not shop. It simply has a want, a wallet and the authority to reorder, within a narrow lane.

Toothpaste, detergent, paper towels, coffee filters, water filters, trash bags, batteries for the smoke alarm, vitamins, pet food. These are the purchases of maintenance, the things a household buys because it ran out, not because it wanted something new. Nobody enjoys buying them. They cost attention, which is exactly what we are trying not to spend.

So they are the first purchases to go to delegates, and they go almost completely. Not because the delegate is brilliant, but because the purchase was never worth a human’s attention in the first place.

What happens to the shelf

For a century, the fight over toothpaste was a fight for the shelf and for your memory. Brands paid for the eye-level slot in the store. They spent fortunes on advertising so that when you stood in the aisle, one name would come to mind first. The whole business of everyday goods was built on influencing a hurried human at the moment of choice.

A delegate is never hurried and has no aisle. It does not remember jingles. It reads the specification: the ingredients, the fluoride content, the price per ounce, the return history, the shipping time, the recall record. It compares every option at once, every time, and it does not get bored and grab the familiar box.

That does not mean brands die. It means the job of a brand changes. A brand used to be a promise made to a person’s memory. For a delegate, a brand has to be a record: a long, checkable history of delivering what it said it would. The brand that wins the toothpaste of 2040 is not the one with the catchiest song. It is the one with the cleanest file.

What the human keeps

There is a twist, and I think it is a happy one. When delegates buy the toothpaste, the choice that is left to a person becomes more meaningful, not less. Most people will never think about toothpaste again. But some people care about it, the way some people care about coffee or knives. For them, the purchase they make by hand becomes a small pleasure, chosen on purpose. The routine goes to the machine. The care stays with us.

The brand used to be a promise made to a person’s memory. For a delegate, a brand has to be a record.

Chapter 8

Batteries

Will delegates buy batteries? Yes, and not only for us. Batteries are where the delegate’s hunger meets the physical world, and they are the clearest answer to anyone who thinks this economy will be weightless.

Delegates with bodies

So far I have mostly talked about delegates that live in software. But a growing share of the trillions will have bodies. A car that drives itself. A robot that cleans, carries, picks or stocks. A drone that delivers. A lawn mower, a vacuum, a security camera, a tractor. Each of these is a consumer in exactly the sense of Chapter 1. It has a want, which is to keep doing its job. It has a wallet, delegated by its owner. And it has the authority to keep itself running.

A machine with a body eats in a way software does not. It needs electricity, delivered to a particular place at a particular time. It wears out. Its tires, motors, bearings, filters, blades and sensors need replacing. It breaks and needs repair. It can cause damage and needs insurance. Every one of those is a purchase, and a machine that can buy will make them itself.

Energy is the delegate’s food

Remember the subsistence basket. Every delegate, with a body or without, eats energy. For software, that energy is bought upstream, by whoever runs the data center. For a machine with a body, the energy is bought at the point of use, a charge here, a charge there, and it is stored in a battery.

That makes the battery the delegate’s stomach. It is how a machine carries its food with it. And a world with billions of machines that move is a world that buys batteries the way the twentieth century bought gasoline: constantly, everywhere, as a basic condition of motion.

The interesting part is who does the buying. A fleet of delivery robots will not wait for a human manager to notice that a pack is degrading. The robot will know its own battery health better than anyone. It will schedule its own swap, compare suppliers, and pay. The battery market of the 2040s will be a market where most of the buyers are the machines that hold the batteries.

Power becomes the new rent

There is a larger version of this. Every delegate in software ultimately runs on electricity from a power plant, through a wire, into a building full of chips. As the trillions grow, the demand for electricity grows with them. Power becomes the thing every delegate must pay for, every second, just to exist.

That is the definition of rent. It is what you pay to occupy the ground you stand on. In the old economy, people paid rent for space. In the delegate economy, the most universal rent is paid for power. Whoever owns the generation, the storage and the wires will collect it from trillions of tenants.

The battery is the delegate’s stomach. It is how a machine carries its food with it.

Chapter 9

Fashion

Will delegates buy fashion? This is the most interesting question on the list, because the answer is no and yes at the same time, and the reason why tells you what fashion has always been.

What fashion is for

In 1899 Thorstein Veblen published The Theory of the Leisure Class, and gave us the phrase conspicuous consumption. His point was that much of what people buy is not for use but for showing. An expensive coat keeps you no warmer than a cheap one. What it does is tell other people something about you: that you can afford it, that you belong to a certain group, that you have a certain taste. Fashion, at its core, is a signal sent to an audience.

A delegate has no body to dress, no mirror and no party to walk into. So in the ordinary sense, no, delegates will not buy fashion for themselves. A machine does not need a coat.

But a delegate absolutely has an audience. Other delegates are watching it. When my delegate approaches a seller’s delegate to make a purchase, the seller’s delegate wants to know: who is this, who does it work for, can it pay, will it keep its word, has it caused trouble before? And my delegate needs to show it can be trusted. That is a signaling problem. It is exactly the problem fashion solves for people.

Fashion for machines

So delegates will buy signals. Not cloth, but credentials.

• Verified identity: proof of which person or business it acts for.
• Bonds and insurance: money posted to show it will make good on a failure.
• Track records: a long, clean history of transactions honored.
• Certifications: third-party attestations that it follows certain rules.
• Scores: reputations computed by networks that watch how delegates behave.

These are the delegate’s wardrobe. They are bought, they are displayed, and they change how others respond. The card networks are already building reputation scores for agents. This will become a large market, and like human fashion it will have its premium brands, its cheap imitations and its fads.

Fashion for people, split in two

Now the human side. Delegates will buy a great deal of clothing for us, and that will split fashion down the middle.

The routine half of a wardrobe is maintenance, not expression: socks, undershirts, the same jeans in the same size, the work shirt you buy every year. That half behaves like toothpaste. It goes to the delegate, it gets bought on specification, and the brands that win it will be the ones with the best record of fit and wear.

The expressive half goes the other way. When a delegate can buy any garment perfectly sensibly, the garment a person chose, by hand, for no reason except that they loved it, becomes a stronger signal than ever. Effort is a signal. Choice is a signal. In a world where the sensible purchase is automatic, the unnecessary one, made in person, is how you show who you are.

I expect a new kind of label to appear on the expensive half of the market, something like chosen by a person. It will sound silly at first. Then it will become the whole point.

Fashion was always a signal sent to an audience. Delegates have an audience too, so they will buy signals: not cloth, but credentials.

Chapter 10

Land

Will delegates buy land? This is the question that sounds most like science fiction, and it has the most down-to-earth answer.

Everything that thinks lives somewhere

A delegate has no house, but its thinking does. Every token is computed on a chip, in a rack, in a building, on a piece of ground, next to a wire that carries power and a cable that carries data. The cloud is not a cloud. It is real estate.

And it is a very particular kind of real estate. For a person, the old rule was location, location, location: near work, near good schools, near the water. For compute, the rule is power, connection and cooling. Ground near abundant electricity, fast fiber and water or cold air is worth far more to the delegate economy than ground with a view.

Who does the buying

Today those sites are chosen by people at large companies, working with brokers, utilities and county planning boards. That will continue for a while, because these are large, slow, political purchases. But look at what goes into the decision: power availability, grid capacity, interconnection queues, water rights, tax incentives, zoning, fiber routes, climate, land price. It is precisely the kind of many-variable comparison that delegates do better than people.

So I expect the search, comparison and negotiation for land to move to delegates long before the signature does. The final step, the deed, will be signed by a human for a long time, because a deed is a promise to a community and a community wants to look someone in the eye. But the delegate will find the parcel, model it, price it and make the offer.

The two land markets

This splits land the way fashion split. There will be land for people: homes, farms, parks, the places we live and love. And there will be land for thinking: parcels whose value comes from what flows through them, the electrons in and the answers out.

I have argued in another series that people will increasingly subscribe to their lives rather than own them, and that the companies offering those subscriptions will become large buyers of land. Put that together with this chapter and you get a picture of the 2050s in which a great deal of land is bought by entities that are not people, chosen by delegates, for uses no person will ever visit. The people who own the ground near the power will be the landlords of the delegate economy.

For a person, the rule was location, location, location. For compute, it is power, connection and cooling.

Chapter 11

Subscriptions

Will delegates buy subscriptions? Here is the answer that surprised me most when I worked it out: they will buy far fewer of them than we do, and they will be the reason humans buy more.

Why subscriptions exist

A subscription is a bundle across time. Instead of paying every time you use something, you pay a flat amount for the right to use it whenever you like. For the seller, it smooths revenue. For the buyer, it does something more important. It removes decisions.

If you had to decide whether to pay every time you played a song, you would play fewer songs and enjoy them less. The meter in your head would spoil it. The subscription takes the meter away. It is, at bottom, a device for saving human attention. We happily overpay for capacity we do not use in exchange for not having to think about it.

Delegates do not need the bundle

A delegate does not have that problem. Deciding costs it almost nothing. It can weigh the price of every call, every query, every second of compute, and never feel the meter. So a delegate does not need to be protected from deciding. Given the choice, it would rather pay precisely for what it uses, because it can see the waste in a flat fee that a person cannot see.

That is why, inside the delegate economy, the meter wins. Pay per call. Pay per query. Pay per second. Pay per result. The tiny payments that were impossible for people, because each one cost more in attention than in money, become the natural way machines pay machines.

Bundle outside, meter inside

Now put the two halves together, because this is one of the clearest pictures of the future I know.

On the outside, facing the human, everything bundles. People will buy fewer, larger, simpler subscriptions: one for their home, their transport, their health, their AI. I have called this the Bundled Economy, and I believe it is where we are headed, because a person’s attention is the scarcest thing they have and the bundle protects it.

On the inside, behind the bundle, everything meters. Trillions of delegates buy and sell from one another in tiny, exact amounts, every second. The subscription you pay once a month is broken into billions of metered purchases the moment it crosses the threshold into the machine world.

The human sees a single price. The delegate sees a river of transactions. Both are right. The bundle is how people want to buy. The meter is how machines want to buy. The economy of 2056 will run on both at once, with the line between them sitting exactly where human attention ends.

Figure · Chapter 11

Bundle outside, meter inside

Outside: what the person sees

• One household plan
• One price a month
• No decisions about each use

The bundle protects human attention.

where attention ends

Inside: what the delegates do

• Power bought by the minute
• Data bought by the call
• Services bought by the result

The meter wastes nothing a machine can see.

The bundle is how people want to buy. The meter is how machines want to buy. The line between them sits exactly where human attention ends.

Chapter 12

What They Will Never Buy

I have spent five chapters on what delegates will buy. It is just as important to say what they will not, because the shape of the hole tells you where the human economy goes.

Things that are only worth having if you were there

Some purchases are not about getting a thing. They are about having an experience, and an experience cannot be delegated without being destroyed.

A delegate can book the restaurant, but it cannot eat the meal. It can buy the concert tickets, but it cannot hear the music. It can arrange the trip, but it cannot see the coast at dusk. It can find the gift, but the gift means something only because you chose to give it. In each case the delegate can do everything except the part that mattered.

I find it helpful to sort every purchase into two piles. The first pile is things you buy so that you do not have to think about them: toothpaste, insurance, storage, power, the routine half of the wardrobe. That pile goes to delegates, and the sooner the better. The second pile is things you buy so that you can think about them, feel them, share them and remember them. That pile stays with you, and it is going to get bigger.

The bakery principle

In the scene in Chapter 17 there is a bakery that is busier than it has ever been. I put it there on purpose. When delegates take the first pile, they free up the one resource the second pile needs: a person’s attention. Hours that used to go to errands, comparisons, renewals and reorders become available for things that are worth being present for.

This is the same thing that happened when Ford gave his workers Saturday. A day off is not just an absence of work. It is time to consume the things that need a person to consume them. Delegates are about to give everyone a great deal of Saturday.

So the businesses that sell the second pile, food made by hand, places worth visiting, teachers worth learning from, craft, care, music, conversation, are not threatened by the trillion-consumer economy. They are its beneficiaries. Every hour a delegate saves is an hour a person can spend on what only a person can enjoy.

The things that must stay human by rule

There is one more category, smaller but important. Some purchases will stay human not because the experience requires it, but because we decide they should. Choosing a doctor for a parent. Hiring the person who will teach your child. Signing the deed to a house. Giving to a cause. Voting, which is not a purchase but is the closest thing a citizen has to spending authority over the whole.

These are matters worthy of human attention. A delegate can prepare them, research them and lay out the choices. But the decision is the point, and a person should make it. I tell my students to delegate the small matters and keep what is worthy of human attention. In a trillion-consumer economy, that advice becomes a description of how the whole economy sorts itself.

Delegates can do everything except the part that mattered.

Chapter 13

You Cannot Advertise to a Delegate

If most of the buyers in 2056 are delegates, the most important question for anyone who sells anything is how to win them. The answer begins with an impossibility.

Persuasion needs a mind to work on

Advertising, as we have known it, works on a hurried human at the moment of choice. It plants a name in memory. It attaches a feeling to a product. It interrupts. It repeats. It borrows the face of someone admired. All of that assumes a buyer whose choices are shaped by memory, mood and attention that can be captured.

A delegate has none of that to work on. It does not remember last night’s commercial. It does not feel anything about a logo. It cannot be interrupted, because it is not looking at anything. It reads what it needs, compares, and chooses. You cannot take an algorithm out to dinner, as one of the first writers on machine customers put it.

The road runs through the record

My own framework, the Reality Equation, has something precise to say here, and it applies to delegates exactly as it applies to people and institutions. Anything that is having a reality, a person, a company, a nation, or a delegate, meets what actually happens against what it expected. Reality is the ratio of the two: R = A / E. The Actual is what the world hands it. The Expectation is what it brings.

The rule I hold most firmly is that no entity can touch either term directly. A seller cannot reach into a buyer’s expectation and set it. Not with an ad, not with a slogan, not with money. The only path runs one way. A seller acts. The action leaves artifacts: deliveries made, refunds honored, specifications met, complaints resolved. Those artifacts become history. History is Actual. And only Actuals retrain the buyer’s prediction.

Advertising never broke that rule. It worked through it. An ad is itself an artifact, a thing that actually happened to a person, and a person’s prediction learns from impressions as well as from deliveries. Seen a hundred times, a name starts to feel like a track record. A delegate does not learn that way. Its prediction of your business is trained on the record of what you delivered. The ad still happens, but it carries nothing the delegate has any reason to learn from.

You cannot advertise to a delegate. You can only leave a record for it to read.

What wins instead

So what does a business do? It stops trying to persuade and starts trying to prove.

• Make the offer readable: what you sell, what it costs, when it is available, and on what terms, in a form a delegate can read without guessing.
• Keep a clean record: every delivery, every return, every promise kept, kept where it can be checked.
• Be payable: accept the ways delegates pay, including tiny amounts and instant settlement.
• Answer the delegate’s questions: let it ask, and answer truthfully and immediately.
• Be the face: keep a human who will stand behind the record when something goes wrong.

Money still matters. But money buys artifacts, not standing. You can spend to make your product better, your delivery faster and your record longer. You cannot spend to skip the record. The road to the delegate’s expectation runs through what actually happened.

The local business

I run a small AI business in Charleston, and most of the businesses I work with are restaurants, hotels and local owners. For one of them, a hibachi restaurant, we built an AI that answers the phone and takes reservations. Today the person on the other end is a human. Before long, more and more of the callers will be delegates, booking a table for someone who never picked up a phone.

That restaurant does not need a bigger ad budget to win those callers. It needs its menu, prices, hours, availability and policies to be readable by a machine, its record of honoring reservations to be clean, and its owner to be a face that people trust when they walk through the door. That is the whole playbook for the trillions, and it is available to a small business on the same terms as a large one. For the first time in a long while, the advantage of the biggest advertising budget is shrinking.

Chapter 14

Money Small Enough to Think With

A consumer needs a wallet. For a person, the wallet is a card, an account, a phone. For a delegate, the wallet has to be something new, because the delegate buys in amounts and at speeds that the old money was never built for.

Money is sized for its buyers

Every form of money in history was shaped by the purchases it had to make. Coins were sized for the market stall. Banknotes were sized for wages and rents. Checks were sized for bills paid once a month. Cards were sized for the shop counter and then the checkout page. Each one carried an assumption about how big a purchase is, how often it happens, and how long a person is willing to wait for it to settle.

Delegates break every one of those assumptions at once. They buy in amounts smaller than any coin, a hundredth of a cent for a piece of data. They buy thousands of times a minute. They buy from sellers on the other side of the world they have never met. And they cannot wait three days for a payment to clear, because the task they are doing will be over in three seconds.

What a delegate’s wallet must do

So money is changing shape again, and it is changing to fit the new buyer. You can derive what it must look like from the needs alone.

• It must handle tiny amounts without the fee swallowing the payment.
• It must settle instantly, so that a seller can deliver the moment it is paid.
• It must carry the mandate with it, so that a seller can check what this delegate is allowed to spend.
• It must leave a receipt a machine can read and a person can audit.
• It must be stoppable, so that a principal can freeze a delegate that goes wrong.

The payment industry is building exactly this, from several directions at once. The card networks are issuing tokens that let a delegate pay with a card under limits the owner sets. Others are using digital dollars that settle in seconds and can move in fractions of a cent. The web’s old payment code is being answered. Which of these wins matters less than what they share. Every one of them is money resized for a buyer smaller and faster than a person.

The allowance

There is a familiar human model for all of this, and I find it clarifying. It is the allowance.

A parent gives a child a fixed amount to spend, with some rules about what it can be spent on, and a way to see where it went. The child learns to make choices inside the limits. The parent stays responsible for the household. The allowance grows as trust grows.

That is what a delegate’s wallet is: an allowance with a mandate attached. My delegates have one. It is small where the risk is high and generous where the work is routine. It grows when a delegate has a long, clean record, and it shrinks when one makes a mistake. In the economy of 2056, I expect most of the money in motion at any moment to be allowance money: authority sent down, spent in small amounts, and reported back up.

Every form of money in history was sized for its buyers. Delegates are the smallest, fastest buyers there have ever been.

Chapter 15

The Mandate

Every delegate spends someone else’s money. That makes the most important document in the delegate economy not the product page or the price list, but the mandate: the grant of authority that tells a delegate what it may buy, for how much, from whom, and when it must ask.

The oldest problem in economics

Economists have studied this relationship for a long time. They call it the principal-agent problem. In 1976 Michael Jensen and William Meckling wrote the paper that made it famous. Whenever one party, the principal, hands authority to another, the agent, to act on its behalf, the agent may not do exactly what the principal would have done. It may be careless. It may be captured by someone else. It may pursue its own interests. The losses from that gap, and the cost of watching and bonding the agent to close it, are called agency costs.

Every company, every government and every household that ever hired help has faced this problem. The delegate economy faces it at a scale no one has seen: trillions of agents, each acting for a principal, each spending money that is not its own.

Delegation runs down

I have a strong view about how this should work, and it comes from the word itself. I call my AI a delegate, not an agent, because to delegate is to send authority down. Delegation only runs one way. I am not the user of my AI. I am its manager. I decide what is worthy of my attention, and I hand the rest down.

A good mandate follows that principle. It is written by the person who is accountable, it states limits plainly, and it reserves the things that must stay human. My delegates may buy tokens, storage and services up to a monthly limit without asking me. They may not sign a contract that commits me to a person. They may not spend on anyone I would want to deal with myself. The more consequential the purchase, the closer it comes back to me.

The signing moment

There is a line in every chain of delegated buying where a recommendation becomes a commitment. Up to that line, the delegate searches, compares, prepares and proposes. At that line, someone signs. I have called this the signing moment, and I think it is where the most important design work of the next twenty years will happen.

For small, routine purchases, the signing moment will move down into the delegate, inside a mandate. For large, rare and personal ones, it will stay with a person, and the delegate’s job will be to make that moment as quick and as well informed as possible. The art of the mandate is putting the line in the right place for each kind of purchase.

Who pays when it goes wrong

The hardest question in the mandate is liability. When a delegate buys the wrong thing, who pays? The principal who gave the authority? The maker of the delegate? The seller who should have noticed? The payment network that carried the money?

Every one of those answers will be tried. My bet is that it settles where it has always settled in human commerce: with the face. Someone has to be accountable, and accountability needs a person who can be found. That is one more reason the human does not disappear from the trillion-consumer economy. The human is the one who answers for it.

Delegation only runs one way. I am not the user of my AI. I am its manager.

Chapter 16

The Economy Turns Inward

Follow the trillions far enough and something strange happens. Most of the buyers are delegates, and most of the sellers are too.

Delegates selling to delegates

When my delegate needs a translation, it may buy it from a translation delegate run by a company I have never heard of. That delegate, to do the job, may buy a glossary from a third delegate and verification from a fourth. Each of them buys compute, power and storage. A single request from me can set off a cascade of purchases among machines, each one tiny, each one real, before an answer comes back.

This is the economy turning inward. The great majority of transactions will happen between delegates, in the middle of chains whose ends are still human. A person sets an intention at one end. A person enjoys the result at the other. In between lies a dense, fast, mostly invisible market of machines buying from machines.

The measurement problem

Our measures of the economy were not built for this. Gross domestic product tries to count final goods and services so that it does not count the same value twice as it passes through a supply chain. That works when supply chains are long and slow and visible. It will strain when a single human request generates a hundred purchases among delegates in two seconds.

I expect a real fight over how to count. Is the compute a delegate uses up a final good or an intermediate one? Is a delegate that runs a person’s affairs part of the household or a small firm? These sound like accounting questions, and they are, but accounting decides what governments see, what they tax and what they think is growing. A country that counts only human purchases will think its economy is shrinking with its population while its actual activity is booming.

A census of realities

There is a deeper way to see this, and it comes back to the Reality Equation. The equation always begins with a question: whose reality? A person has one. So does a family, a company, a nation. The history of civilization can be read as a census of those realities, from the tribe to the family to the individual, and onward to minds beyond number.

Each delegate with a want, a wallet and a mandate is one more thing meeting what actually happens against what it expected. The number of those is about to rise faster than at any time in history. That is the real meaning of trillions of consumers. It is not just more buying. It is more of the world being met, priced and decided, by more kinds of minds, than ever before.

And at the center of all of it, still, is the person. Not because the person is needed for every step, but because the person is where the want began, and where the result is finally lived.

A person sets an intention at one end. A person enjoys the result at the other. In between lies a dense, fast, mostly invisible market of machines.

Chapter 17

A Morning in 2056

She does not see them unless she asks, and she rarely asks. Her delegates bought power for the house at three in the morning when it was cheapest, and sold some back at six when the neighbors woke up. They reordered the toothpaste and the dish soap. They renewed the car’s insurance for the day, because it would be driving her mother to the clinic. They bought a weather feed, a traffic model, a few seconds of a better translation service for a message to a cousin abroad, and a review of a contract she will sign this afternoon, from a delegate run by a law firm she has never visited.

One purchase was different. Her delegate paid a woman in town, a retired engineer, for forty minutes of her judgment on a renovation plan. The delegate could have produced a plan on its own. It wanted someone who would put her name on it.

Her town is smaller than it was when she was born. There are fewer children in the school and more houses on the market. The economists on the news call it the long decline. But the shops on Main Street are busy, because the delegates send the routine orders to the warehouse and the people come in person for the things they care about. The bakery is the busiest it has ever been. Nobody’s delegate buys their bread.

At the edge of town, where the old mill used to be, a long low building hums next to the substation. No one works there, and no one visits. It is one of the busiest places in the county.

She finishes her coffee. Her calendar shows one appointment today that her delegates could not take for her. It is lunch with her son.

What is different, and what is not

Look at what changed between my morning in 2026 and hers in 2056. The number of purchases made on her behalf went up by a factor of a thousand. The number made by her own hand went down. The share of her attention spent on buying anything at all fell almost to zero.

And look at what did not change. She still wants things. She still decides what matters. She still pays, through one line, for everything. Her town has fewer people and more consumers. The bread is still bought by hand. The lunch is still hers.

That is the picture I believe in. Not a world where machines replace people as consumers, but one where the count of consumers detaches from the count of people, and the people move to the top of the hierarchy, where their attention is spent only on what is worthy of it.

Chapter 18

What to Do Now

A book about 2056 is only useful if it changes something you do in 2026. So here is what I would do, depending on who you are.

If you are a person with a household

Start your short list. Write down the purchases you never want to think about again and the ones you always want to make yourself. The first list is what your delegates will take. The second is what you are protecting.

Write your first mandate. It does not have to be elaborate. A sentence will do: what your AI may buy without asking, up to how much, and what it must always bring to you. Then actually let it. The point is not the money saved. It is learning, early and at low stakes, how to be a good manager of something that spends on your behalf.

Read your receipts. Not every one, every day. But once a month, read the long list the way I read mine, and ask whether each line was worth what it cost. That habit is the beginning of what I call being a Super Admin: someone who runs their AI rather than being run by it.

If you run a business

Make yourself readable. Your menu, your prices, your hours, your availability, your policies and your way to pay should be available in a form a machine can read without guessing. If a delegate cannot read your offer, you are invisible to a growing share of buyers, and you will never know what you lost.

Keep a clean record and keep it where it can be checked. Every promise kept is now a form of advertising, the only form a delegate can see.

Be payable by a machine. Make sure the ways you accept money include the ways delegates pay.

And be the face. Somewhere in your business there needs to be a person your customers trust, someone who answers when the record is not enough. That person is your most valuable asset in the trillion-consumer economy, because that person is the one thing your competitors’ delegates cannot copy.

If you are a student

I have advised my students, who are about twenty-four, to let their first employer be their first customer: to sell results rather than hours, and to keep their own costs low so that they are free to choose. This book adds one more piece to that advice.

Learn to sell your attention to machines. The delegates of the next thirty years will need people who can judge, witness, sign, teach and show up. Become someone whose judgment is worth paying for and whose name is worth putting on the work. That is the top shelf of the trillion-consumer economy, and it is open to anyone willing to be accountable.

Be someone whose judgment is worth paying for and whose name is worth putting on the work.

Epilogue

The Receipt, Again

I used to read the long list with a little worry, the way you read a teenager’s credit card bill. Now I read it the way a farmer reads a field. Each line is a small piece of work getting done that I did not have to do. Each one was a decision made by something that works for me, inside limits I set, on behalf of things I care about.

The short list is still mine. Groceries. Gas. A dinner out. A gift. Those I bought by hand, on purpose, because they were worth my attention.

The world is about to fill with consumers who are not people. There will be trillions of them by the time the human count starts to fall. They will buy almost all the toothpaste, all the power, most of the batteries, the land near the wires, and a river of tiny services by the meter. They will buy credentials instead of clothes. And at the very top of their list, the thing they cannot make and will pay most for, they will buy us.

The count of heads will go down. The count of consumers will keep climbing. And the short list, the one you make by hand, will be the most valuable list in the economy.

Keep it short. Make it yours.