The Leisure That Never Came

Keynes famously predicted in 1930 that his grandchildren would work 15-hour weeks. In his essay Economic Possibilities for our Grandchildren, he reasoned simply: if output per worker kept compounding the way it had, humanity would eventually produce all it needed in a fraction of the time, and the old “economic problem” — the struggle for subsistence — would largely be solved. Work would shrink dramatically, perhaps to three-hour shifts, leaving people with a problem that must have sounded almost luxurious during the Great Depression: what do you do with all that free time?

Nearly a century later, his productivity forecast looks remarkably prescient. Output per worker has multiplied, machines have absorbed enormous amounts of physical labour, computers have compressed tasks that once took days into seconds, and entire categories of work have become unimaginably easier. What did not arrive was anything resembling the leisure Keynes imagined. Working hours did decline substantially through parts of the nineteenth and twentieth centuries, but nothing close to the trajectory implied by a 15-hour week. Instead, much of our additional capacity became more output, higher incomes, cheaper and better products, new industries, greater profits and a constantly expanding definition of what constitutes a normal standard of living.

We became vastly richer. We did not become correspondingly more idle.

Perhaps the mistake was assuming that productivity gains naturally become leisure. They do not. Productivity creates a surplus, and what happens to that surplus is a separate question.

Why Efficiency Rarely Means Doing Less

There is an old economic observation that helps explain part of this. In the nineteenth century, William Stanley Jevons noticed something puzzling while studying coal use in Britain. Steam engines were becoming more efficient, meaning less coal was required to perform a given amount of work. You might therefore expect total coal consumption to fall. Instead, it rose.

Greater efficiency had made steam power cheaper, which made more uses of it economically worthwhile. Once coal could produce more work for the same cost, activities that had previously been uneconomic became feasible. The improvement did not merely reduce the coal required for an existing set of tasks; it expanded the number of tasks worth doing.

Something similar seems to happen, in a looser sense, whenever productive capacity becomes dramatically cheaper.

Imagine that a piece of software once took a team of developers twelve months to build, and new tools reduce that effort to two months. One possibility is obvious: build the same software and spend the remaining ten months doing something else. In practice, that is rarely what happens. The company adds features that were previously too expensive, launches another product, serves a smaller market that could not previously justify the cost, runs more experiments, supports more edge cases and moves faster because competitors are moving faster too.

The productivity gain does not simply shrink the old workload. It changes what becomes worth doing.

Once every competitor has access to the same improvement, what initially looked extraordinary can quickly become the new baseline. If one company uses a tenfold productivity improvement to dramatically reduce effort while another uses it to launch more products, respond faster and lower prices, the first company may discover that its newly won leisure carries a competitive cost.

This is why the seemingly obvious equation — twice the productivity should mean half the work — rarely survives contact with an actual economy. Greater efficiency changes not only the cost of doing the things we already wanted to do, but also the universe of things worth attempting in the first place.

Keynes May Also Have Underestimated “Enough”

There was another assumption hidden inside Keynes's forecast: that human wants would eventually reach something like saturation. Once people had enough food, shelter, clothing and material comfort, perhaps the attraction of earning still more would diminish. Leisure would become more valuable than additional consumption, and people would naturally choose to work less.

Looking forward from 1930, this was not an absurd assumption. Looking backward from today, it seems much less convincing.

As societies became richer, the definition of a decent life expanded with them. Once food and shelter were secure, there were better homes, safer neighbourhoods, cars, air conditioning, better healthcare, university education, foreign travel, restaurants, entertainment, retirement savings, smartphones, streaming subscriptions and experiences that did not even exist when the previous definition of “enough” was formed.

Many of these things genuinely improved life. The point is not that we should have refused them in favour of afternoons in a hammock. It is that human wants proved far more elastic than a model based on fixed material needs would suggest. Each generation inherited things the previous generation would have considered luxuries, then quietly incorporated them into its definition of ordinary life.

There is also another class of things that productivity can never fully satisfy because their scarcity is relative rather than material. Everyone can live in a larger home than their grandparents did, but everyone cannot live in the most desirable neighbourhood. Everyone can become richer in absolute terms, but everyone cannot earn more than their peers. The same applies to career position, reputation, university places, influence and market share.

If everyone in an industry becomes twice as productive while success is still measured partly by outperforming others, there is no obvious reason for everyone to work half as much. The race can simply re-anchor at a faster pace.

This is where Keynes's apparently simple progression — more productivity, satisfied needs, less work — begins to come apart. Productivity increased enormously, but so did expectations, consumption and the things against which people measured a successful life.

“Enough” kept moving.

Productivity Creates the Possibility of Leisure, Not Leisure Itself

There is another part of this story that is easy to oversimplify. It is tempting to say that working hours should automatically have fallen as productivity increased, but employers or owners of capital simply captured the gains. Bargaining power certainly matters, and the shorter working days and weeks that emerged across the nineteenth and twentieth centuries did not materialise merely because machines became more efficient. They reflected worker pressure, regulation, rising incomes, employer competition, changing labour markets and changing social expectations.

That distinction matters because there is no rule saying that an hour saved by technology must return to the person whose work became more efficient. The hour might become higher wages, lower prices, greater profits, more output, better quality, another customer or another product. It might simply become a new expectation about how quickly something should now be done. It can also become leisure, but leisure is one possible destination for the surplus rather than its automatic destination.

Seen this way, Keynes's mistake was not really about the machines. He understood remarkably well what technological progress could make possible. What he seems to have underestimated was how institutions, markets and human preferences would respond once those possibilities arrived.

We Do It to Ourselves Too

It is easy to tell this entirely as a story about companies capturing productivity gains, but there is a smaller and perhaps more revealing version happening in ordinary life.

Email was faster than letters, but we did not respond by spending dramatically less time communicating. We began sending and receiving far more messages. Search engines made information vastly easier to find, but we did not simply answer the same number of questions more quickly; we started researching questions that previously would not have justified the effort. Word processors made editing almost frictionless, but documents did not necessarily become quicker to finish. They acquired more revisions. Messaging tools removed much of the friction from workplace communication, and communication became continuous.

Faster tools often do not merely compress an existing activity. They expand it.

AI may make this pattern unusually visible because the time savings can be so dramatic. Something that once required three hours can suddenly take thirty minutes. But what happens to the remaining two and a half hours? Often we ask for another version, explore another possibility, analyse more information, polish the output further or start a task that previously would not have been worth starting.

Sometimes nobody takes the saved time away from us. We spend it ourselves.

For centuries, we have invented machines partly because they promise to save us time, yet whenever we successfully save some, we seem remarkably capable of discovering more things worth doing with it. The washing machine does not necessarily create an empty afternoon; other activities expand around the time it releases. Faster transportation does not necessarily mean less time travelling; it can mean living farther away. Faster communication does not necessarily mean fewer minutes spent communicating; it can mean communicating with more people, more often.

Capacity expands, and life reorganises itself around the new capacity.

Almost a Century Later, the Same Prediction

This is where the story starts to feel strangely familiar.

It is now 2026, and much of the conversation around AI sounds remarkably like a prediction made in 1930. Machines will handle more of the work. Human labour will become dramatically less necessary. Working weeks will shrink. Work may eventually become optional. Some form of universal basic income may distribute the abundance generated by automation, leaving people with far more time than previous generations could imagine.

And that is the more interesting déjà vu. We do not need to retrofit Keynes's “grandchildren” into the year 2030 to see it. Nearly a century after he wrote that technological progress would finally release people from much of the burden of work, we find ourselves making essentially the same prediction about the next technological leap.

Perhaps this time really is different. AI may be capable of automating broad categories of cognitive work in a way that previous technologies were not, and genuine discontinuities do happen. The failure of earlier predictions does not prove that this one must fail too.

But Keynes gives us a useful reason to separate two questions that are often treated as though they were the same.

• CanCan technology dramatically reduce the amount of human labour required to produce what we need? History gives us every reason to think that it can.
• WillWill we therefore choose to work dramatically less? History is much less reassuring.

The first question is technological. The second involves everything technology cannot settle for us: what we want, what we regard as enough, how strongly we compete for relative position, how productivity gains are distributed, what institutions reward and what sort of life we consider worth living.

That was the part Keynes seems to have misread. His machines were not the problem. The extraordinary productivity gains arrived. What did not follow automatically was the human decision to translate those gains into less work.

And now, in 2026, another generation is looking at another astonishing machine and telling itself that this one may finally give us back our time.

The Time We Save

Maybe it will.

But if history is any guide, the interesting question is not how much time AI can save us.

It is what we will find to do with the time once it does.