The CEO of ChatGPT Says Your Workweek Is About to Change
Sam Altman published a 13-page policy document calling for a four-day workweek. You can disagree with every proposal in it and still need to run the audit it implies: which day of your current week disappears if the hours shrink?
By Forge Team
If your employer handed you a four-day week tomorrow, which day would you give up? Most professionals have never had to answer that question, so the honest answer is "I'm not sure — probably a bad one." That gap is the practical problem worth solving this month, no matter what policy ever gets passed.
On April 6, Sam Altman published a 13-page policy document titled "Industrial Policy for the Intelligence Age." Axios reported the headline proposals: a public wealth fund seeded by AI companies, taxes on robot labour, an expanded four-day workweek with no pay loss, a "Right to AI" for every citizen, and directed investment in human-centred sectors like care work and education. Altman is explicit that "widespread job loss" is a near-term consequence of the technology his company is selling.
You can disagree with every proposal in that document and still find the underlying signal useful. The CEO of the world's largest AI company is saying — publicly, in policy language — that the shape of the work week is the thing about to change. That is not a forecast. It is an instruction to start running the experiment inside your own calendar, now.
The four-day thought experiment
Take last week. Block out Friday. Now look at the remaining four days and ask: what did you do on Friday that actually mattered, and what did you do because it was on the list? Most professionals find the first list is shorter than they expect and the second is longer.
Picture a product manager running this exercise. Her Friday last week included three status updates, two competitor scans, a deck polish for Monday, and one call that ended in an actual product decision. The call is the only thing she could not plausibly hand off to a current model. Everything else is a candidate.
A mid-level lawyer running the same exercise finds her Fridays are mostly first-pass contract review and the tail end of client email — both activities where current AI already produces output she has to correct rather than write from scratch. The cost of having AI do the first pass is the cost of supervising it, which is not zero but is lower than starting from blank.
A marketing manager's Friday list is briefings, template updates, and review cycles. AI can carry two out of three. The third — the briefing session with her design team — stays, because that is where judgement compounds into something specific to her product.
The thought experiment is not about whether the four-day week is coming. It is about learning to tell the difference between the task you do because no one else can, and the task you do because no one has questioned it yet.
Which of your weekly tasks are genuinely AI-suited? 90 seconds to find out.
The part Altman is not saying out loud
A four-day week sounds like a gift. In practice it is a forcing function. Somebody — probably you — has to decide which tasks vanish, which get handed to AI, and which stay human. If you do not decide, someone with less context will.
The risk inside the gift is that professionals under time pressure hand over the wrong tasks. The tempting ones to delegate are the hard ones, because hard tasks hurt. But hard tasks are usually hard because they involve judgement, and judgement is the part of your role that is still scarce. Once you delegate it, you stop practising it. Once you stop practising it, you forget how to tell good output from bad.
The HR analyst who lets AI draft every round of performance feedback is not saving time. She is losing the annual reps that taught her how to read a team. The account executive who lets AI write every client follow-up is not freeing up hours. He is losing the read on each client that made him closeable in the first place.
There is a version of the four-day week where AI absorbs the mechanical work and professionals spend the recovered hours on the judgement work they used to skip. There is also a version where the mechanical work is gone, the judgement work has atrophied, and nobody notices until a senior decision goes visibly wrong. The difference between the two versions is decided at the individual level, in exactly the audit this post is describing.
Which tasks should you refuse to hand off?
Three moves before the end of the month
One. Run the Friday audit. Not as a thought experiment — on paper, for last week and this week. Label every task "delegate-ready," "delegate-with-oversight," or "keep human."
Two. For every task in the "keep human" column, write one sentence on why. If you cannot, it probably belongs in another column. If you can, you have the language to defend that time the next time your manager asks where it went.
Three. Pick one judgement-heavy task you quietly handed to AI in the past six months and start doing it again, by hand, once a week. That is your insurance against atrophy.
Pick the judgement work you will keep doing by hand.
Altman's document is not a prediction you need to agree with. It is a prompt to answer a question your manager will eventually ask and your employer will eventually act on: if your week shrinks, what survives? Running the audit now is free. Running it after somebody else has decided for you is not.
Put this into practice
Reading is a start — but skill comes from doing. Try these drills now.
Like this post?
Get the next one in your inbox. Practical AI skills, no filler.