How AI Is Making the Four-Day Work Week Possible

The largest UK pilot held, and most companies kept it.

The UK pilot ran from June to December 2022 with 61 organisations and just under 3,000 workers, each committing to a 20% cut in hours at full pay. It was coordinated by the Autonomy Institute with 4 Day Week Global, and researched by the University of Cambridge, the University of Salford and Boston College. UKRI's summary of the ESRC-funded research reports:

• 56 of the 61 companies were still operating a four-day week a year later
• A 65% reduction in sick days, and a 57% fall in the number of staff leaving
• Revenue barely changed — up 1.4% on average across the 23 organisations that provided data
• 90% of employees wanted to continue; more than two thirds reported lower burnout

Read that carefully and the headline is not productivity. It is that output held while the cost of churn and absence fell. The hours came out of somewhere, and that somewhere is the subject of this post.

In the work that is repetitive, rule-bound and currently done by a person.

Not in the thinking, the selling or the judgement. The fifth day is mostly made of tasks nobody would defend in a meeting:

• Enquiry triage — reading, categorising and routing what arrives
• Drafting — first versions of replies, proposals, summaries and updates
• Document handling — pulling structured data out of invoices, forms and contracts
• Reporting — assembling the same numbers into the same shape every week
• Scheduling and chasing — the follow-ups that exist only because something is pending

Ranked by how fast they pay back, triage and document extraction come first — the buyer's guide to AI services sets out the order and why.

These are the same starting points that make AI pay for a small business at all — the case for removing repetitive work rather than replacing people is identical; the shorter week is simply what you do with the hours once they exist.

Because a four-day week fails on coordination, not on capacity.

Remove ten hours of admin from a team that has no documented workflows and you get a team with ten free hours and the same bottlenecks. The day disappears back into meetings and waiting.

What actually breaks on a shorter week is the invisible coordination that a fifth day used to absorb: the quick question, the handover nobody wrote down, the approval that only one person could give.

So the automation is necessary and not sufficient. The operating model is the other half.

Four rules, each of which is boring and load-bearing.

• Document every workflow. If it lives in one person's head, their day off is an outage
• Hand off asynchronously. Written status by default, so nobody waits on a conversation
• Automate with an approval gate. Anything leaving the business gets a human yes
• Name one owner per workflow. Not a team — a person, with a named deputy

That is the same governance we apply to any automation we ship, and it is not a coincidence. A shorter week and a trustworthy automation need the same thing: work that can run without the person who built it standing next to it.

Monday to Thursday, with clients across seven markets.

We work a four-day week — Monday to Thursday, 9:00 am to 6:00 pm. It is on our Google Business Profile and it is not a perk we hand out quietly; it is the operating constraint the rest of the business is built around.

Covering seven markets on four days works because of the rules above rather than in spite of them. Client-facing commitments are scheduled inside the four days. Anything that would need a Friday is either automated, handed to a named deputy, or honestly scheduled for Monday rather than promised and missed.

The constraint is useful. A week with no slack exposes bad process immediately, which is the opposite of comfortable and the reason it works.

Anywhere the service is the coverage.

Be honest about the exceptions rather than pretending the model is universal:

• On-call and incident response — someone has to be reachable; rotate it, do not pretend
• Live client-facing coverage — support desks and sales floors need staffing, not shorter weeks
• Shift and physical work — hours tied to a place and a machine do not compress

For those, the realistic version is staggered four-day weeks across a rota rather than a closed Friday. The business still gets the retention and absence benefits; it just does not get a quiet office.

Three to six months, measured on output, with the automation done first.

• Automate before you shorten. Take the repetitive hours out first, or you are just asking for more in less time
• Agree what output means — the one or two numbers the team is accountable for, written down before you start
• Run three to six months. Shorter than that and you are measuring novelty
• Measure absence and retention too — in the UK pilot these moved further than revenue did

The competitive argument is worth stating plainly: a smaller firm that runs this well keeps people that larger competitors lose, which is one of the few places where being small is an advantage rather than a handicap.

If you are buying this rather than building it, the engagement should end with a working system and a handover — that is what an AI consultant is for.

AiS LABZ builds the automation half of this — applied AI and automation, with an owner, an approval gate and an audit log on everything that ships. Founded in 2024 by Salman Usman, we start with a free diagnostic.