If AI does all the work, who buys the products?

Of all the objections raised against the automated economy, this one is the favourite of practical people. Fine, the machines make everything. Machines do not buy sneakers. If nobody earns a wage, who exactly is the customer?

It usually gets asked as a gotcha, proof that the whole thing cannot happen. It deserves better than that, because it is pointing at something true and important. It just points at the wrong conclusion.

Why the question is better than it sounds

Step back and look at what a wage actually does in an economy. It pays you for your work, obviously. But it also does a second job that nobody thinks about: it is the pipe through which the money from production flows back out to the people who buy the products. People are paid to make things, and their pay buys the things. Demand and supply are two ends of one circuit, and for the last two centuries the wage has been the pipe connecting them.

So the practical people are right about the mechanics. A company can automate away its own workers, but it cannot automate away its own customers. Every firm that cuts its wage bill is quietly assuming that everyone else keeps paying wages, so that somebody out there can still afford its products. If they all do it together, the assumption fails together. An economy that lets pay collapse without building a replacement is sawing through the branch its own profits are sitting on.

Where the gotcha goes wrong is in assuming the pipe must be a wage. What the circuit needs is that households hold a claim on what production earns. A wage is one kind of claim. It is not the only kind.

The pipe is already narrowing

This is not a thought experiment about some future robot economy. The pipe has been narrowing for forty years.

Across the fifty-nine countries where it can be measured, the share of national income flowing to people who work for a living has been falling since the early 1980s, in forty-two of them. And you can watch what that does to the customer side of the circuit in the American spending data: the top tenth of earners now account for 49.2 per cent of all consumer spending, the highest share since records began in 1989, up from about 35 per cent in the early 1990s. The bottom 80 per cent of households have merely kept pace with inflation since the pandemic. Half the customer base of the world's largest consumer economy is now one household in ten.

That is what an economy looks like while its main demand pipe slowly closes: it does not stop, it concentrates. The products keep selling, to fewer and fewer people, who own more and more of the machines. AI does not create this problem. It finishes it, faster, which is also this book's answer to most questions of the kind.

The other pipe

There is a second way money reaches a household, familiar to anyone who owns shares: the dividend. If you own a piece of the productive economy, you get paid when it earns, whether or not you clocked in.

That is the direction the whole question points once you take it seriously. If automation shrinks the wage pipe, the circuit can only keep running if the ownership pipe widens: if ordinary households hold claims on the machines, so that what the machines earn flows back out as spending power. The profits fund the dividend, the dividend funds the buying, and the buying sustains the profits. A closed loop, yes. But every economy is a closed loop. The only question is which pipe carries the flow, and who is standing at the tap.

This is not utopian plumbing. One place has run a small version for four decades: Alaska pays every resident an annual dividend from a fund built on the state's oil wealth, and has done so through every crash since 1982. In 2008, the year the financial system nearly failed, the payout was the highest the programme had ever made. The cheque shrank the following year and the system carried on, because the fund was designed to smooth bad years rather than pass them straight through to households.

The honest reading

Two concessions, so nobody thinks the loop is magic. First, a fund holding shares in the machines will crash when markets crash. That is survivable with boring, tested rules: spend less than you earn, average the payouts over good and bad years. The mechanics have their own chapter in the book. A dividend has bad years the way wages have recessions. Second, wages are not about to vanish. Most people will earn pay for a long time yet. The argument is about direction: every year the wage pipe carries a little less of the flow, and the ownership pipe carries a little more, to the small group already holding it.

Which is why the timing matters more than anything else in this essay. Spreading ownership is easy to arrange while the machines are still being built and the wealth is still forming; it is brutally hard after, when it means taking things from people who have them. Wait too long and you are trying to prime a pump that has already run dry.

So: who buys the products? Whoever holds a claim on what the machines earn. Today that is a narrowing circle of owners, and the spending data shows the circle tightening year by year. The question was never whether there will be customers. It is who gets to be one. Not what the machine can make. Who owns the machine?

---

Where the evidence lives. The treadmill between capital and wages is chapter 5 of The Horse Is Here to Stay; what happens to a citizens' dividend when markets crash is chapter 10. The running record of events is kept in the Stable.

One of a series of free guides answering the questions people are actually asking about AI and work. The book carries the evidence, the sources and the counter-arguments in full. These guides are published under a Creative Commons Attribution-ShareAlike licence: translate them, teach from them, quote them, build on them.