If AI Makes Us More Productive, Why Are We Still Working Five Days?

The Monday-to-Friday week feels permanent, but it was negotiated rather than inevitable. The move from six-day weeks towards shorter hours took decades of campaigning, legislation and changes in how businesses operated, and Henry Ford's introduction of a five-day, 40-hour week for many of his factory workers in 1926 became one of its best-known milestones.

What happened next is more revealing. Computers replaced filing cabinets, email transformed communication and software automated countless administrative processes, yet none of these office technologies shortened the standard working week. The gains went into more output, faster service and new kinds of work. AI differs in reach rather than principle, because it targets the drafting, summarising and analysis that fill so much of an office worker's day. Whether it breaks the pattern depends on whether anyone decides it should, and that's the question the rest of this article examines.

The phrase "four-day week" covers several very different arrangements, and the numbers make the differences clear. Imagine you earn £40,000 a year for a 40-hour week. If you move to 32 hours with a matching 20% pay cut, you'll earn £32,000, and losing £8,000 of gross income would rule the arrangement out for many people. If instead you squeeze your 40 hours into four ten-hour days, your salary stays the same but you're working exactly the same hours. You've changed your schedule rather than your working time.

The version most people actually want is 32 hours on the same £40,000. This is often described as the 100:80:100 model: 100% of pay for 80% of the time, in return for a commitment to maintain 100% of productivity. From the employer's point of view, though, the hourly cost of your labour has just risen by 25%, so the model only works if output genuinely holds up.

Here's where AI could change the calculation. Take a hypothetical team of ten people, each working 40 hours, giving 400 hours a week in total. The business introduces AI tools that help with routine reports, produce meeting notes and make information quicker to find across several systems. After measuring the results carefully, it finds it can maintain the same service and output with 320 hours instead of 400. If that improvement is real and sustained, the same ten people could each work 32 hours, keep their salaries and still deliver what the business needs.

That's an illustration rather than a prediction, and real life is messier. AI tools cost money, some work still needs human review, customer demand doesn't arrive neatly within shorter hours, and saving time on individual tasks rarely translates into saving the same proportion of a whole working week. The possibility still matters, because it suggests AI could make shorter hours affordable without making lower salaries necessary.

If AI is going to make shorter weeks affordable, the obvious place to start is the evidence from organisations that tried them without it. In 2022, before generative AI tools were common at work, 61 UK organisations employing around 2,900 people took part in a six-month trial of reduced working hours with no loss of pay. The pilot was organised by 4 Day Week Global with the think tank Autonomy and the 4 Day Week Campaign, with researchers from the University of Cambridge, Boston College and University College Dublin.^[1]

The results were striking. Of the 61 organisations, 56 said they would continue with the shorter week after the trial, and 18 confirmed the change as permanent straight away. Some 71% of employees reported lower levels of burnout and 39% said they felt less stressed. Among the organisations that provided comparable financial data, revenue stayed broadly level over the trial period, rising by 1.4% on average when weighted by company size. A follow-up published in 2024 found that at least 54 of the 61 were still running a four-day week a year later, and at least 31 had made it permanent.^[2] It's one thing to experiment for six months and quite another to decide the change is worth keeping.

The limitations matter. The organisations volunteered, so they weren't a representative sample of British employers. Maintaining revenue isn't the same as maintaining profit, and results varied between businesses. But the trial began in June 2022, months before ChatGPT was released, and it showed that some organisations could keep performing with fewer hours even without today's AI tools. That doesn't prove AI enables a four-day week. It does make it reasonable to ask whether combining the lessons of these pilots with genuinely useful automation could make shorter weeks workable for organisations that couldn't previously make the numbers add up. We don't yet have the evidence to answer that confidently, but it's a question worth testing properly.

There's a problem with assuming that time saved by AI turns into free time. If AI cuts the time needed to prepare a report from three hours to thirty minutes, your manager is unlikely to suggest you go home early. You'll probably move on to the next task, prepare another report or find that reports are now expected much faster. None of that is necessarily bad, since businesses exist to serve customers and grow. But making individual tasks faster is a different thing from reducing the hours people are expected to work.

The research supports this. As I discussed in the first article in this series, a large study of more than 7,000 knowledge workers found that people with access to AI tools saved around two hours a week on email, but the overall shape of their work didn't change.^[3] A shorter week would need someone to make a deliberate decision, based on measuring where time is really being saved, checking that output and service quality hold up, and formally changing how work is organised. Without that, we could spend the next decade celebrating enormous productivity gains while working exactly the same hours, possibly under even higher expectations.

There's another outcome that worries me. Imagine your employer announces that everyone can finish on Thursdays, but your meetings, deadlines and customers' expectations stay exactly as they were. Before long you're skipping lunch, working later and checking messages on your supposed day off. Technically you have a three-day weekend, but in practice you're exhausted by Thursday evening.

AI could help prevent this by removing routine tasks and unnecessary administration. It could equally make things worse if employers respond by continually raising expectations. If AI helps you complete ten tasks instead of eight, it's a short step to asking for twelve, and once twelve is normal, fifteen. There's no natural point at which an organisation has to declare that enough work has been done. A genuine four-day week has to be designed around a sustainable workload rather than a more demanding performance target, and it needs clear boundaries about whether people are expected to answer urgent messages on their day off, whether they can properly disconnect and whether AI is reducing the effort involved or simply increasing the volume of work passing through. Otherwise, we've changed the calendar without improving anyone's working life.

It's relatively easy to see how AI could make a shorter week possible for someone who spends most of the day working with information. Reports, emails, research, analysis and administration all contain tasks AI can speed up. The picture is very different for a nurse, a chef, a delivery driver or a roofer. A patient still needs care, a restaurant still needs staff on shift and a roof still needs fixing. AI might improve scheduling, reduce paperwork or help prepare a quote, but it rarely reduces the hours someone needs to be physically present.

That creates a real cost problem. If a hospital moved nurses from five shifts to four on full pay while patient care still required the same cover, it would need to recruit more staff. A small restaurant can't shorten its opening hours because its booking system has become more efficient, and a tradesperson can't finish a physically demanding job 20% faster because AI helped write the estimate. Shorter weeks may still be achievable in these sectors, but they're likely to need staggered rotas, additional staff or different ways of organising services, all of which cost money.

There's a fairness question here that deserves more attention than it gets. We could easily end up with office workers enjoying four-day weeks on full pay while people in some of the most physically demanding and socially essential jobs continue working five. The people whose work AI can speed up most easily aren't necessarily the people who would benefit most from an extra day off.

If you wanted to propose a four-day week on full pay to your employer, supported by the AI tools your team already uses, the argument that you'd enjoy having Fridays off is unlikely to win. Your employer would want evidence: which tasks genuinely take less time, how much is really saved once checking and correcting AI output is included, whether service and quality can be maintained, whether colleagues would need to cover the extra day, what it would cost and whether it would last beyond a short trial. Employees have equally reasonable questions about whether the workload would stay manageable, whether the day off would be protected, how pensions, holidays and career progression would be treated, and whether the option would be open to every team.

A sensible starting point is a carefully designed pilot. Measure the team's current output, hours, costs and wellbeing, introduce or improve AI-supported processes where they genuinely help, and then test whether contracted hours can be reduced while maintaining service quality and financial performance. Above all, don't assume that faster individual tasks automatically justify a 20% cut in everyone's hours. Measure what actually happens.

It's also worth knowing the legal position. In the UK, employees have had a statutory right to request flexible working from their first day of employment since April 2024^[4], but that isn't an entitlement to a 32-hour week on full pay, and employers don't have to approve every request. Measured evidence of what AI actually saves is what turns a request into a business case. That's probably how shorter weeks will spread, through individual organisations showing that a different arrangement works rather than through a single dramatic announcement.