The Decision Nobody's Making When AI Frees Up Your Time

If AI has made you noticeably faster this year, you've probably not decided yet what fills the time it freed up. That's not a lapse in discipline. It's the one decision almost every AI-adoption story on record skips — and skipping it is why the fastest-growing group of AI users isn't the happiest one.

The pattern across four signals

Rachitsky's second tech-worker survey split the workforce three ways: 49% feel their output amplified by AI, 14% feel destabilised by it, 37% sit in between. The difference wasn't the tools everyone had access to — it was whether they'd built a repeatable workflow with a review step in it, instead of improvising a new prompt every time.

His first survey, two weeks earlier, had already shown the cost of skipping that step: 82% of respondents said AI made them more productive, and burnout rose from 44.7% to 55.7% year over year. The gap wasn't people getting slower. Nobody had decided where the recovered time should go, so requests filled the space automatically.

Anthropic's Claude Reflect dashboard, launched the same month, named a version of the same gap from the product side. It asks users what they want to keep doing themselves — because the company building the tool could see people quietly losing track of which tasks they were still capable of checking closely.

And the earliest piece in this set, on why productivity and exhaustion arrive together, explains the mechanism: AI removes the low-demand work that used to double as recovery time, and replaces it with continuous evaluation — judgement calls with no natural stopping point, which is more tiring per hour even when it produces more.

Four sources, four different months, the same underlying shape. AI creates capacity — in time, in attention, in output. By default, that capacity gets absorbed by whoever or whatever moves first: a manager's rising expectations, a habit of not looking too closely, more work sliding into the gap. Rarely is it a decision anyone made on purpose.

What to do differently on Monday

The fix isn't "use AI less," applied as a general rule. It's specific: before you start a task with AI, decide three things — what level of review it gets, how many rounds it's allowed, and where the time you save is going. Skip that decision and one of the outcomes above makes it for you instead.

Tomás — the manager who didn't ask

Tomás runs a five-person analytics team at a 90-person logistics company. Over six months, AI cut his team's report turnaround roughly in half. Nobody on the team had a conversation about what to do with the freed-up hours — so his own manager did, by routing two new client accounts to the team at the old staffing level. Output stayed high. Everyone on the team, Tomás included, was now doing the work of what used to be seven people, and nobody had agreed to that.

Renata — the one who set the terms first

Renata is a grants writer at a regional nonprofit. When her organisation adopted AI for first drafts, she set a rule before writing a single proposal with it: two revision rounds per draft, and any time saved under her old average stayed logged as hers, not reassigned. Her director agreed in writing. Eight months in, she's producing more proposals in the same hours she worked before — and she still catches subtle errors in AI drafts, because checking for them stayed a deliberate step instead of something she stopped doing once the drafts looked polished.

The one thing

Two surveys, one product feature, and one burnout mechanism are describing the same missing step: deciding on purpose what happens to the capacity AI creates, before habit, a manager, or a deadline decides for you.