Four hours in seventy-five years: why AI productivity will not buy you a Friday

The pilots and the history are not in conflict; they measure different things. The 4 Day Week Global trial ran 33 businesses and more than 900 workers for six months on full pay for 80% of the hours, and the firms that supplied numbers reported an 8% revenue rise during the trial and 38% against the same period a year earlier ^[6][7]. That is a self-selected group of employers that chose to run the experiment, reporting against its own past, over half a year. The 38-to-34-hour move since 1950 is every American employer over three quarters of a century ^[2]. One is a demonstration that the arrangement can work. The other is a record of what happens when nobody is running a demonstration.

The Fortune author's own 2018 survey of 3,000 employees across eight countries found 45% thought they could finish their tasks in five hours a day without interruptions, while 49% of US respondents said they regularly worked overtime ^[4][5]. Read those two numbers together and the constraint is not the volume of work. If half of workers already believe the job fits in 25 hours and work more than 40 anyway, the missing hours are being spent on visibility and on the friction a long week creates [c5a]. An AI tool that removes an hour of that friction does not produce an hour of leisure; it produces an hour that still has to be visibly occupied.

Which is why the strongest number in the pilot literature is the one about money. In the 4 Day Week Global trial, 42% of employees said they would need a 26% to 50% raise to go back to five days, and 13% said no sum would do it ^[8]. Workers who have held the shorter week price it at roughly a third of their salary. Nobody in the AI-abundance conversation is proposing to hand over a third of payroll in time. Mark Dixon of IWG, which the piece describes as the largest flexible workspace provider with more than 8 million users across 122 countries, told Fortune the answer on shorter weeks is "no time soon" because everyone is focused on productivity, and because firms cannot pay the same wages for fewer hours or pass the difference to customers ^[10][11].

That is the mechanism the predictions skip. Jamie Dimon has forecast a four or three-and-a-half-day week within a few decades, Anthony Scaramucci a three or four-day week within our lifetimes, and Bill Gates has floated a two-day week on AI abundance ^[9]. All three treat hours as the residual once productivity rises. The author's counter-example is concrete: the PC extended the workday into evenings, and the internet followed workers home ^[12]. Microsoft Japan's much-cited 40% productivity gain came with capped meetings, a five-person attendance limit, 23% lower electricity costs and 60% less printing ^[3]. Those are the marks of a deliberately redesigned week, not of a technology that shortened one on its own.

The demand side is not the obstacle either. Asked how many days they would work at constant pay, only 4% of respondents in the 2018 study said zero, and the largest group, 34%, chose four days, ahead of the 28% who chose five ^[13]. Given the choice, most people pick a shorter week and keep working. What they do not control is who books the hour that a model saves.