AI will not give you Fridays off

John Maynard Keynes once imagined a future where rising productivity would cut the workweek to 15 hours. That didn’t happen. In 1950, Americans averaged 38 hours a week. Today, the average is 34. A century of electrification, PCs, the internet, and cloud computing shaved off four hours, not three days.

That matters because the latest round of AI cheerleading is making the same promise all over again. Jamie Dimon has talked about a four- or three-and-a-half-day week within a few decades. Anthony Scaramucci sees a three- or four-day week in our lifetimes. Bill Gates has floated a two-day week. The optimism is understandable. The history is not on their side.

The evidence for shorter weeks is strong, just not the route to getting there. Microsoft Japan said its productivity jumped 40% after moving to a four-day schedule, while meetings were limited to 30 minutes and five people. Electricity costs fell 23%, and the company printed 60% fewer pages. In a larger trial run by 4 Day Week Global with more than 900 workers across 33 businesses, people kept 100% of their salary while working 80% of the time. Workers rated it 9.1 out of 10, 97% wanted to continue, and none of the participating companies planned to end it. Reported revenue rose 8% during the trial and 38% versus the same period a year earlier.

But the problem isn’t whether shorter weeks can work. It’s whether the system lets the gains reach workers. The piece argues that technology tends to expand output, not leisure: the PC stretched work into evenings, and the internet followed people home. Mark Dixon of IWG put it bluntly: everyone is focused on productivity, so not soon. In other words, AI is likely to make people faster at their jobs, then hand them more jobs.

That is the real hinge. The U.S. still treats 40 hours as the full-time norm, and there’s no near-term federal push to change that. AI may save workers more than two hours a day, but unless labor rules and company incentives change, that time will be filled, not freed. Keynes missed that part. So do a lot of AI optimists.