What the hype says
In recent months Jamie Dimon, Anthony Scaramucci and Bill Gates have suggested that artificial intelligence could soon deliver a three-day or even a two-day workweek. Their optimism rests on the idea that AI will automate routine tasks and free up large blocks of time for employees.
Why the promise matters
Productivity has risen dramatically since John Maynard Keynes imagined a 15-hour week in 1930. In the United States the average workweek fell from 38 hours in 1950 to 34 hours today, a modest four-hour reduction despite electrification, personal computers, the internet and cloud computing.
Research I have overseen for seven years, covering pilots in Iceland, Japan, the United Kingdom, Australia and the United States, shows that many workers feel they could complete their duties in five hours a day. In a 2018 survey of 3,000 employees across eight countries, 45% said they could finish tasks in five hours without interruption, yet 49% of American respondents reported regularly working overtime.
Trials that have reduced the workweek do show benefits. Microsoft Japan reported a 40% rise in productivity after moving to a four-day schedule, cutting meetings to 30 minutes and limiting attendees to five. The company also saw electricity use drop by 23% and paper consumption fall by 60%.
A large experiment coordinated by the nonprofit 4 Day Week Global, with partners at Boston College, Cambridge University and University College Dublin, placed more than 900 workers in 33 firms on a four-day week for six months while keeping full pay. Participants rated the experience 9.1 out of 10, 97% wanted it to continue and none of the firms planned to end the policy. Reported revenue rose 8% during the trial and 38% compared with the same period the previous year. Burnout fell and productivity rose, while 42% said they would need a 26-50% pay rise to return to a five-day week.
Why the evidence has not moved the needle
The obstacle is not technology but the economic system. In the United States the standard 40-hour week has been set by the Fair Labor Standards Act since 1940 and there is no current federal legislation encouraging shorter weeks. Companies facing cost-of-living pressures are reluctant to pay the same wages for fewer hours, and they rarely pass efficiency gains on to employees.
"Everyone is focused on productivity, so no time soon," said Mark Dixon, chief executive of flexible workspace provider IWG.
Dixon argues that each wave of technology, from the PC to the internet, has expanded the amount of work rather than reduced it. The same pattern is expected with AI: faster development will create more tasks, not fewer.
What could change the outcome
Successful policy experiments, such as the Icelandic reform that now gives 86% of the workforce reduced hours or the right to request them, show that legislation can shift the balance. Without similar legal frameworks, the calculations offered by Dimon, Scaramucci or Gates do not hold in the American context.
My own research with GoTo found that AI tools can free more than two hours per day for many workers, roughly ten hours a week. In practice, organisations tend to allocate that extra capacity to additional projects rather than to shorter schedules.
To translate efficiency into leisure, companies would need to redesign work structures, unions would have to negotiate reduced hours, and governments would need to create incentives or mandates. Until those systemic changes occur, AI is likely to keep the workweek where it is.
Conclusion
AI will make employees more efficient, but history suggests the extra time will be filled with new work unless policy and culture evolve. The four-day week is not a technological problem; it is a labour-policy challenge.