Workers now spend 6 hours a week babysitting AI. The frustration isn't the bot — it's the slop
A new figure says office workers burn six-plus hours a week "botsitting" — supervising, correcting and cleaning up after AI. The thread of practitioners largely shrugged at the babysitting itself. What actually grinds them down is reviewing un-checked AI output a colleague forwarded without reading.
One manager sent a developer an AI-generated spec to rebuild a core feature the team had shipped two years ago. The frustration, as one commenter put it, isn't disrespect for you — it's disrespect for the work.
On the productivity claim, the most credible voices land near 20% self-reported gains, with a tiny group seeing 2–3x — and a 4,000-team study suggesting output isn't the same as value.
Six hours of supervising a machine
The report puts a number on a feeling: workers spend over six hours a week "botsitting" — the hidden human labor of prompting, checking and fixing AI output — and it's fueling workplace frustration. The thread, mostly people who do this daily, didn't dispute the hours. They relocated the grievance.
The problem isn't the bot. It's the un-read handoff
The most-upvoted reframe drew a clean line. A commenter separated the two things the article blurs:
"I could care less about bot sitting (haven't we always written our own automation?), but it's botsitting the unverified slop that people send you that fuels frustration. I thought I worked with competent people who respected me."
The concrete case followed immediately. A developer described his product lead sending "an AI generated PRD (complete with a Claude Code spec!) to build a core feature which we have had for over 2 years." A reply found the precise nerve:
"It's not a lack of respect for you; it's a lack of respect for the work itself. That lack is being rewarded and encouraged."
What the productivity actually measures
The thread's calmest analyst tried to size the real gain. A commenter who tracks this closely gave a range, not a slogan:
"GenAI is a self reported a 20% efficiency boost, and for a very (very) small group of people, it's maybe a 2-3x boost… for most use cases, AI productivity is either the equivalent of giving people 3D printers, and seeing little benefit, or signing up for an outsourcing service."
A reply split the measurement: a first-draft boost is real, but "6 hours of debugging and docs reading is not equal to 6 hours of prompt fiddling." For harder evidence, a commenter pointed to an observational study of 4,000 teams over two years — direct measurement from version control and CI, not a survey — to make the point that "productivity != value."
The slot machine
The most honest description of the daily experience came from someone who enjoys it and resents it at once. A commenter compared botsitting to reverse-engineering a black box:
"I kind of enjoy exploring black boxes, trying how different inputs are mapping to differences in outputs. It's kind of like hacking. The problem is, they keep altering the box."
That's the trap in one line. The skill you build supervising a model is non-transferable, because the model under you changes every few weeks. Six hours a week isn't the cost people resent. It's six hours spent re-learning a box that someone else keeps swapping out — and then being handed a coworker's un-read output to debug on top.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
Solar out-generated coal in the US for the first time. An ex-nuclear engineer added the asterisk
For the first time, solar produced more US electricity than coal in a month. It's a genuine milestone — and the thread, full of energy professionals, immediately put it in proportion: gas still out-produces solar nearly five to one, and coal's collapse is mostly a switch to gas.