The promise is a shorter workweek

Technology leaders have repeatedly argued that AI will automate enough work to make shorter schedules possible. The BBC notes that OpenAI urged companies to test a four-day workweek, even though a former employee said the lab had not trialled that schedule during their time there.

The prediction can still become true. It is not a technological consequence. A company must decide to convert higher output per hour into fewer hours without cutting pay, rather than raising expectations or reducing headcount.

The reported reality is a permanent sprint

A former OpenAI technical employee told the BBC they worked at least 70 hours a week. Other workers said sprints at OpenAI and Anthropic can continue for weeks and exceed 90 hours across seven days. At Meta, current and former employees described urgent transfers into AI work, late nights, weekends, and an on-call feeling outside formal working hours.

These accounts cannot establish an industry-wide average. They do reveal the incentives at the frontier, where competitive deadlines and enormous investment can turn every efficiency gain into pressure to ship faster.

AI can save time and expand the job

Research cited by the BBC followed hundreds of technology workers using AI and found faster pace, broader task scope, and work extending into more hours. Workers also spent time checking outputs and adapting processes around the tools.

That helps explain why measured task savings may coexist with worse work-life balance. New capabilities create new requests, new supervision, new integration work, and a new baseline for what one employee is expected to deliver.

Publish the productivity bargain

Companies claiming AI-driven productivity should disclose more than adoption and output. They should track hours, after-hours contact, task expansion, rework, health indicators, staffing, wages, and how much saved time employees actually retain.

The central question is not whether AI can make a task faster. It is whether the institution lets people live any part of the future that the technology was supposed to create.

  • Measure total hours and after-hours availability alongside AI adoption.
  • Separate task speed from rework, supervision, and newly created work.
  • Set limits on prolonged sprints and involuntary transfers into urgent AI projects.
  • Share productivity gains through time, pay, staffing, and worker voice.
Primary trail

Go to the source

Read the evidence behind this analysis. External links open in a new tab.

BBC — Tech leaders promise less work while staff report extreme hours Harvard Business Review — AI can intensify rather than reduce work