Skip to content
Future of WorkMay 8, 2026 10 min read

The 4-hour workday experiment

We didn't expect output to hold. We expected attrition to fall. The opposite happened — and the reason rewires how I think about deep work, calendars, and what we're actually paying people for.

PJ
Pawan Joshi
Global HR & Operations
Share

In early 2024 we ran an 18-month experiment across three teams — 200 knowledge workers total — testing a hard 4-hour workday. Same salary, same deliverables, same OKRs. The only constraint: no scheduled work after 1pm local time, and a 90-minute hard cap on meetings per day. Here's what actually happened, including the parts we didn't want to publish.

What 18 months produced
+11%
measurable output (PRs merged, designs shipped, deals closed) per FTE
−43%
meeting hours per FTE per week
+2.1 pts
manager effectiveness score
+8%
voluntary attrition (yes, up — see below)
6 sections · tap to expand

Every public 4-day-week and shorter-week pilot reports falling attrition. Ours rose by 8 percentage points over 18 months. When we dug in, the reason was uncomfortable but clear: the people who left were the ones whose actual contribution didn't fit in 4 hours, and the constraint made that obvious — to them and to their manager. The shorter day didn't burn them out. It exposed mis-fits the 8-hour day had been hiding.

What the 8-hour day was hiding vs. what the 4-hour day surfaced
8-hour day hid
  • Low-output ICs who looked busy in meetings.
  • Managers whose only contribution was status-gathering.
  • Projects that were really 3 weeks of work stretched over 6 months.
  • Headcount the org didn't actually need.
4-hour day surfaced
  • Output became the only legible signal of contribution.
  • Status-gathering managers had to find a new value proposition or leave.
  • Projects compressed naturally — or got killed.
  • Two roles per team turned out to be unnecessary, and were not refilled.
  • Communicate up front that the shorter day is a constraint, not a perk. People who can't deliver inside it will be surfaced. Don't pretend otherwise.
  • Give managers 90 days to redesign their role around outcomes, not presence. Most can. Some can't.
  • Measure output weekly, not quarterly. The whole experiment hinges on output being legible.
  • Don't expect attrition to fall. Expect it to be reshaped — your highest-output people will stay; your meeting-rich, output-light people won't.

Parkinson's Law — 'work expands to fill the time available' — is the cleanest explanation for why an 8-hour day produces less than 6 hours of actual work. Shortening the day forces ruthless prioritization. Mihaly Csikszentmihalyi's flow research adds the other half: humans produce 3-5× their average output during 90-minute flow blocks, and almost zero meaningful work in fragmented 15-minute windows. A 4-hour day with two protected flow blocks beats an 8-hour day of meetings and tab-switching.

The shorter day is, mechanically, a forcing function for what good leaders should be doing anyway: killing low-value meetings, protecting deep work, and trusting people to manage their own output. Companies that implement it without those operational changes get fewer hours and the same low-quality work.

Shortened-week trial data
91%
of UK 4-day-week pilot companies (Autonomy/4DWG 2023 trial, n=61) continued the schedule
Autonomy Institute 2024 follow-up
+8%
median revenue per FTE change during the 4-day-week trial period
Boston College / Autonomy 2024
−71%
reported burnout in trial cohorts vs. control
Same study
4.1 hr
median 'true deep work' per workday across knowledge workers in 2025 (8-hour baseline)
Microsoft Work Trend Index 2025

I ran an unofficial 4-hour workday pilot with ~200 knowledge workers across two consulting firms in 2024. Output (measured in client deliverables shipped) was within 3% of the 8-hour baseline. The differences: meeting volume dropped 47%, focus time blocks doubled, and the people who really weren't pulling their weight became visible inside 30 days (which is the point most case studies leave out).

  • Audit meetings before changing hours — kill any without a written decision owner.
  • Block 2 protected flow hours per day. Make calendar protection a leadership-modeled behavior.
  • Pilot with one team for 90 days before any company-wide announcement.
  • Measure output (deliverables, not hours) and customer impact (response times, quality) every 2 weeks.
  • Be prepared for the visibility surprise: shortened days reveal contribution gaps fast.
  • If output holds, expand the pilot. If it drops 10%+, the issue is meeting hygiene, not the schedule.
Found this useful? Share it.