AI will not give you Fridays off | Fortune

In 1930, John Maynard Keynes wrote one of the most optimistic predictions in the history of economics. Thanks to technology and rising productivity, he argued, future generations would work no more than 15 hours a week. They would have so much leisure time they would barely know what to do with it. Keynes was one of the most brilliant economic minds of the twentieth century. He was also spectacularly wrong on this question. In 1950, Americans averaged 38 hours a week. Today, we average 34. Nearly a century of technological progress, from electrification to the PC to the internet to the cloud, moved the needle by four hours.

I have been researching the four-day workweek for seven years, tracking pilots across Iceland, Japan, the UK, Australia, and the U.S. In 2018, I led one of the largest global studies on working hours conducted at the time, surveying 3,000 employees across eight countries including the United States, Britain, and Germany. What we found was telling: 45% of workers believed they could easily finish their tasks in five hours a day without interruptions, but many were exceeding 40 hours a week anyway, with the United States leading the way, where 49% said they regularly worked overtime. Workers were not burning through those extra hours because the work demanded it. They were filling time, managing appearances, and absorbing the inefficiencies that long workweeks encourage.

So when I heard Jamie Dimon predict that AI will usher in a four or three-and-a-half-day workweek within a few decades, or Anthony Scaramucci declare we are moving to a three or four-day week in our lifetimes, or Bill Gates float the idea of a two-day workweek powered by AI abundance, I understand the optimism. I also understand, based on seven years of data and a century of history, why it is almost certainly wrong.

The Evidence for Shorter Workweeks Is Strong. The Path to Getting There Is Not.

Let me be clear about something: a shorter workweek can benefit both companies and employees. The data supporting it is among the most consistent in modern workplace research. Microsoft Japan reported a 40% productivity boost after moving employees to a four-day schedule. Meetings were capped at 30 minutes. Attendance was limited to five people. Electricity costs fell 23%. The company printed 60% fewer pages. Less time, more focus, better results.

A landmark trial coordinated by nonprofit organization  4 Day Week Global, with research partners at Boston College, Cambridge University, and University College Dublin, put more than 900 workers across 33 businesses on a four-day schedule for six months, paying them 100% of their salary for 80% of the time. The results were striking: workers rated the experience 9.1 out of 10, 97% said they wanted to continue, and not a single participating company planned to discontinue the policy. Businesses that provided data reported an 8% revenue increase during the trial period and a 38% increase compared to the same period the prior year. Self-reported burnout and fatigue declined, productivity went up, and 42% of employees said they would need a 26% to 50% pay raise to return to a five-day week. Thirteen percent said no amount of money would get them back.

My own 2018 research found that only 4% of workers, when asked how many days they would want to work if pay remained constant, said zero. The biggest share, 34%, chose four days. The standard five-day week came in second at 28%. People want to work. They just do not want to waste time doing it.

The evidence is not the problem. The problem is the system that evidence must survive in.

Here is the uncomfortable truth about new technologies and productivity revolutions: the gains tend to go to companies, not to workers’ calendars. The PC did not shorten the workweek. It extended the workday into evenings and weekends. The internet did not free us from the office. It followed us home.

Mark Dixon, CEO of IWG, the world’s largest flexible workspace provider with more than 8 million users across 122 countries, said it plainly when asked about Gates’ and Musk’s predictions: ‘Everyone is focused on productivity, so no time soon.’ His reasoning cuts to the core of the issue. Companies and workers are both squeezed by cost-of-living and cost-of-operating crises. Businesses cannot afford to pay the same wages for fewer hours, and they cannot pass the difference on to customers. So any time freed by AI is far more likely to be filled with new tasks than handed back as a long weekend.

Dixon’s broader argument is one I find historically compelling. Every major technological shift has followed the same arc: fear of displacement, followed by an expansion of opportunity and, critically, an expansion of workload. AI will speed up companies’ development, he says, so there will be more work. Just different work. The Luddites smashed looms in 19th century Britain to stop automation. What they got instead was the Industrial Revolution.

The four-day workweek is not a technology problem. It is a policy problem. For example, Iceland ran one of the most successful trials ever documented, and as a result 86% of the country’s workforce now work reduced hours or gained the right to do so. 

However, in the US, the federal standard for a full-time workweek, 40 hours, has not changed since the Fair Labor Standards Act was amended in 1940. No federal legislation mandating or incentivizing a four-day week is on the near-term horizon.

Without that foundation, the math Dimon and Scaramucci are describing simply does not hold in the American context. A CEO optimistic about AI’s long-term impact on working hours is making a prediction about technology. The actual outcome depends on labor law, union density, corporate incentive structures, and the balance of power between employers and employees. Technology is the least complicated variable in that equation.

AI is already saving workers real time. My firm’s research, conducted with GoTo, found that employees are recouping more than two hours per day thanks to AI tools, or over 10 hours a week. That is a meaningful efficiency gain. The question is what happens to those two hours.

Based on everything I have seen over fifteen years of workforce research, the answer in most American workplaces is: more work gets added. When a company discovers that its employees can now process three reports where they previously processed two, the response is rarely to send them home an hour early. The response is to assign a fourth report. 

Dixon put it well when he said AI will speed up companies’ development and therefore create more work, not less. My research found something similar: workers who are already exceeding 40 hours a week are not doing so because they lack the tools to finish faster. They are doing so because the culture, the expectations, and the incentive structures of their organizations reward presence and output volume over focus and recovery.

AI will make workers more efficient. Companies will use that efficiency to do more. And the workweek will stay roughly where it is, because it has stayed roughly where it is through every previous wave of technology that was supposed to free us.

None of this means the four-day workweek is impossible. The research says it works. The pilots say it works. My own data from 2018 says workers want it and are fully capable of delivering it. What it requires is not a better AI model. It requires companies willing to redesign how work is structured, governments willing to create the policy frameworks that make shorter hours viable at scale, and a cultural shift away from the idea that time at a desk is the same thing as value created.

Keynes was not wrong about productivity. He was wrong about what we would do with it. So far, we have done the same thing with every efficiency gain technology has ever delivered: we have used it to do more. Until something changes about the system around the technology, AI will be no different.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.