When Machines Do the Work, What Will We Live On?

The abundance AI creates, and income beyond paid work

October 18, 2026 · Joo-Ho Lee

If we have built technology that needs less human labor, we should also rethink the social arrangements that require us to work long hours to enjoy its benefits.

1. What if there is plenty to buy, but no one can afford it?

Imagine a factory in a city somewhere. Until last year, ten people made 1,000 components a day. This year, AI plans the production schedule, while robotic arms handle assembly and inspection. Now two people watch the monitors as the factory turns out 1,500 components a day. Defects are down, and the cost of making each component has fallen by more than half. For management and shareholders, the numbers tell a wonderful success story.

Now pull the camera back a little. Similar changes are taking place in the other factories down the street, and in warehouses, call centers, and accounting firms in the next city. The people leaving those workplaces were also buying the goods and services they produced until yesterday. The goods may be cheaper and even better in quality. But if new jobs or other sources of income do not fill the gap, where will people find the money to buy them?

Business owners have faced this question before. In 1914, Ford introduced a five-dollar day, including a conditional profit-sharing payment. That was more than double the previous daily wage of $2.34. The familiar explanation is that Ford wanted workers to be able to buy the cars they made. The more immediate problem, though, was high staff turnover: workers kept quitting because they could not stand the monotony of the assembly line. Higher wages can improve productivity and increase spending. But the extra wages a company pays do not all come back to it as sales. Employees buy other companies' products, too. Purchasing power is a question for society as a whole, not just for any one business.

We need to distinguish between two kinds of abundance. One is abundance in production: the ability to make more with fewer people and less time. The other is abundance in everyday life: ordinary people actually enjoying what is produced and living without constant insecurity. We have tended to assume that more of the first will bring more of the second. AI is putting that assumption to the test again, on a larger scale.

2. Does a growing economy mean a better life for us?

The Stanford economist Charles Jones has highlighted a remarkable regularity in economic growth. Over roughly 150 years, from the late nineteenth century to today, US GDP per person has grown at about 2 percent a year along its long-term trend, despite major economic fluctuations. Two percent may sound small, but it means income doubles roughly every 35 years. Compared with our great-grandparents' generation, we live in a world that is several times richer.

Jones emphasizes new ideas as a driving force of growth. Yet when population growth slows and the expansion of the research workforce faces limits, we cannot take continued growth at the present pace for granted. This is where AI enters the picture. If AI takes on some of the work of research and discovery, it could ease those constraints and accelerate growth. That is one reason economists are interested in its potential.

But we need to be clear about something. GDP growth tells us about the size of the pie, not who gets how much of it. An average does not describe everyone's experience. If a billionaire walks into a neighborhood restaurant, the average wealth of its diners may suddenly reach hundreds of millions of dollars. No one else has become any richer.

The data show this gap, too. According to the US Economic Policy Institute, net labor productivity in the United States rose by 59.7 percent between 1979 and 2019, while real hourly compensation for the median worker rose by 15.8 percent. Compensation includes benefits as well as wages. The size of the gap depends on which workers and price measures we use. Still, it is clear that we must separately ask how much of a productivity gain reaches ordinary workers. Research by Loukas Karabarbounis and Brent Neiman also reported a fall of about five percentage points in labor's share of income across the global corporate sector between 1975 and 2012.

As labor's share falls, capital's share rises. The problem is that ownership of machines, software, and shares in companies is highly concentrated. The US Federal Reserve's Distributional Financial Accounts likewise show that most household holdings of corporate equities and mutual funds belong to the wealthiest 10 percent, while the bottom 50 percent hold very little.

AI and robots are part of that capital. When a factory's workforce falls from ten to two, the wages of the other eight do not simply become profit. Machines cost money to buy and run, and strong competition may pass cost savings on to consumers through lower prices. Even so, if the benefits of automation are concentrated among owners and there are few ways to replace lost earnings, the economy can become richer while many people's lives become less secure. Growth alone does not settle the question of distribution.

3. Will new jobs come to the rescue again?

The fear that technology will take away jobs is nothing new. For several years beginning in 1811, textile workers in parts of Britain destroyed textile machinery in what became known as the Luddite movement. Yet jobs did not disappear over the next two centuries. Their number grew.

Agriculture is perhaps the clearest example. Farms were once a major source of American employment. Today, a small workforce produces far more food. The descendants of those who left farming work in factories, offices, hospitals, and software companies. According to research led by David Autor at MIT, about 60 percent of US employment in 2018 was in types of work that had first appeared in occupational classifications after 1940. This does not mean that 60 percent of broad professions were entirely new. But it does show that technology has created new work while eliminating old work.

Will that happen again? It may. But the range of work AI can change is broad. Tractors and looms mainly replaced particular tasks; AI is used across writing, analysis, design, and learning. Even when new work emerges, AI may soon be able to do some of it. We do not yet know how quickly, or in what numbers, the next jobs for people will appear.

We can already see AI helping people do their work. A study of 5,172 customer-support agents found that those using an AI assistant were, on average, 15 percent more productive. The gains were around 30 percent for less experienced or less skilled workers. Higher productivity does not, however, automatically mean fewer jobs. If prices fall and demand grows, employment may hold steady or rise. The important question is who benefits when the same work takes less human labor, and in what form.

The changes are not confined to a computer screen. Daron Acemoglu and Pascual Restrepo studied the adoption of industrial robots in the United States between 1990 and 2007. Their estimates for the country as a whole linked one additional robot per 1,000 workers to a decline of about 0.2 percentage points in the employment-to-population ratio and about 0.42 percent in wages. These findings concern industrial robots in a particular period, not a direct forecast of what future AI will do. Robots with sensors and vision systems existed before, too. Now advances in AI are opening up the possibility of robots working in a wider range of environments. We should pay attention to the prospect of change accelerating both in offices and on the ground.

Humans are not the only ones technology has displaced. As cars and tractors spread, the number of horses used for transport and farming fell sharply. There were not enough "new jobs" for horses. I am not suggesting that people will meet the same fate. People can learn, organize, and change institutions. But should we continue to stake our livelihoods on the assumption that enough new jobs will appear, soon enough, for everyone? Is it still best to depend mainly on employment for income in an age when employment may become less certain?

4. Why not work less and live better?

Let us turn the question around for a moment. Why did we build machines? To lift heavy things for us, go into dangerous places for us, and do tedious calculations for us. If the purpose of technology was to lighten our burden, a world in which people need to work less would be an achievement, not a failure. Why, then, does the prospect make us more frightened than happy?

What frightens us is not less work itself. It is a system in which losing work can also mean losing the foundations of a decent life. Losing a job puts income and a sense of belonging at risk. In a society such as the United States, where health insurance is often tied to employment, access to care can become less secure as well. Under those arrangements, a machine that relieves us of work feels more like a threat than a blessing. The problem lies in the way we have tied our livelihoods to technology and work.

We have, in fact, traveled part of this road already. The long working days and six-day weeks of early industrialization gradually shortened in many countries. Ford introduced a five-day, 40-hour week in 1926. In the United States, the federal threshold for overtime pay fell to 40 hours a week in 1940. That did not impose a blanket ban on working longer than 40 hours, but it helped change society's expectations about working time. The weekend did not fall from the sky. It grew out of efforts and institutional changes that returned some of the benefits of higher productivity in the form of time.

In 1930, John Maynard Keynes imagined that living standards might rise so much over the next hundred years that people would need to work only 15 hours a week. With 2030 only a few years away, his expectations for prosperity and working time have taken different paths. Living standards have risen enormously, and working hours have fallen, but a 15-hour week has not become the norm. We have turned productivity gains into both consumption and leisure, but not in equal proportions, and not with equal benefits for everyone.

Now imagine a different picture. A society where working three or four days a week is enough to stop worrying about rent, food, and medical bills. Where taking time to raise children or care for aging parents is an ordinary part of life, rather than a "career break." Where people can study a new subject after forty, pick up an instrument for the first time at fifty, or spend time solving neighborhood problems with their neighbors. Where giving an exhausted body enough rest is a right essential to a healthy life, not a sign of laziness.

Caring, learning, creating, resting. Paid services count toward GDP, but the value of caring for family at home, studying on our own, or taking a rest is not fully reflected in it. Yet these are precisely the things that shape our quality of life. In a world where machines work for us, time may be the most precious resource left to people. Abundance can mean having more of our own time, not just having more things.

5. Will profit-seeking companies share that abundance?

Think back to the future we imagined as children. Robots cooked, cleaned, and ran factories while people played, learned, rested, and lived comfortably. That was the future in comics and science magazines. But one question was missing from the dream: who owned the robots? As children, we never doubted that the benefits of their work would somehow become ours.

In reality, most robots and AI systems are owned or operated by businesses that seek a profit. Companies mainly adopt machines to cut costs and increase profits, not to give people more time off. That is not an accusation; it is how companies work. So when machines take over work, the freedom not to work does not automatically reach everyone. For some, the first consequence is the insecurity of losing a job.

Return to the factory in the opening scene. If competitors lower their costs through automation, it is hard for one company to carry on as before. Automating is a rational choice for that business. But if many companies reduce employment at once, and lower prices, new jobs, investment, or income transfers do not sufficiently replace lost purchasing power, demand across the market may weaken. Success for individual companies does not lead directly to stability for society as a whole.

Companies' goodwill alone is unlikely to bridge this gap. We need rules that go beyond individual firms. There is no settled answer, but the discussions and experiments so far point to three broad possibilities.

First, share the gains through taxes and public services

Some of the profits generated by AI and automation could be taxed to provide everyone with a basic income, or to guarantee basic services such as healthcare, education, housing, and care.

Cash-transfer experiments show both possibilities and limits. In a three-year US trial by OpenResearch, 1,000 people with low incomes received $1,000 a month. According to the study's August 2026 revision, recipients' labor-force participation rate was 4.2 percentage points lower than that of the control group, and they worked about one to two fewer hours a week. The results suggest that cash can expand people's choices while also reducing labor supply. But a temporary experiment involving a particular group cannot tell us everything about a permanent, tax-funded program covering an entire population.

Second, let citizens own productive assets and receive dividends

Instead of redistributing the proceeds later through taxes, citizens could share ownership of machines and companies from the outset. If income from labor declines, the idea is to spread income from capital more widely.

Alaska offers a real example. Part of its oil revenue has been invested in a permanent fund, which has paid annual dividends to eligible residents since 1982. Studying this program, Damon Jones and Ioana Marinescu found no significant decline in overall employment, while part-time employment increased. That does not mean a dividend of this size could cover all living expenses. The point is that a system for sharing returns from collectively held assets already exists. We can also consider ways to broaden ownership of productive assets and share their returns in the AI era.

Third, share productivity gains through shorter working hours

Here, higher productivity comes back to people as time rather than money: fewer hours of work for the same pay.

In 2022, 61 companies and about 2,900 employees joined a four-day-week pilot in the United Kingdom. After six months, 56 companies decided to continue, including 18 that made the change permanent. Employee burnout fell, and revenue broadly held up among the 23 companies that supplied revenue data. But the participating companies had volunteered, so the findings cannot simply be applied to every industry. Sharing the gains as time also needs to be tested against the realities of different industries and ways of organizing work.

None of these paths is easy

All three pose difficulties. We have to fund them without undermining the incentive to innovate, and account for the fact that capital moves across borders. Giving every adult in the United States $1,000 a month would mean gross payments of roughly $3 trillion a year. The actual net fiscal cost would depend on how much was recovered through taxes and how existing programs changed, but it would not be a small decision.

It also matters who decides how public funds are invested and distributed, and who holds those decision-makers to account. Simply moving concentrated power from companies to the state would not be enough. What to share, how much, and with whom are ultimately questions of politics and institutions. These difficulties are reasons for careful design rather than for giving up. Combining the three approaches to suit each society's circumstances is probably more realistic than choosing one as the answer.

6. Beyond a choice between capitalism and communism

Bring up these ideas, and a familiar response soon follows: "Isn't that just communism?" The question carries the weight of the twentieth century. Markets or planning. Private ownership or state ownership. Capitalism or communism. We are used to sorting the world into two boxes.

But real economies have never fitted so neatly into either one. Alaska is one of the most conservative US states, yet it has shared dividends from common assets with eligible residents for more than forty years. Norway has a market economy and owns one of the world's largest sovereign wealth funds. Most advanced countries today are mixed economies, combining markets and competition with public healthcare, pensions, and unemployment insurance. We have been living somewhere between the two boxes for a long time.

So we should change the question. Instead of asking which system will win, can we preserve the energy of markets and competition without making people's survival depend on wages alone? This is not a declaration of victory for any one ideology. It is the fairly straightforward idea that, as the way we produce things changes, we can rethink the rules of ownership and distribution to match.

Looking back, we have done that whenever production has changed. When land was the source of wealth, who owned it shaped society. When factories became a source of wealth, labor laws, trade unions, the 40-hour week, and social insurance emerged. None of those arrangements was obvious from the start. They came through fierce arguments and resistance. AI and robots are now becoming new sources of wealth. We do not yet know what rules will fit them. But we should not put off the search.

The question is not only how much of our work machines will take over. It is how we can share in the abundance they create by doing that work for us.

A final note

I began writing this after reading an article by a colleague in economics, "Is the Future of AI and Economic Growth Really So Rosy?" and thinking through the question it raised. My own reflections took a very different direction, but that article was unmistakably the starting point. I am grateful to my colleague for giving me such a worthwhile question to think about.

Sources and further reading

Jones · Growth and AI · EPI · Productivity and pay · Labor’s share of income · MIT · New work · Generative AI at Work · Robots and Jobs · OpenResearch · Cash-transfer trial · Alaska dividend study · Cambridge · Four-day week