The ongoing human work of supervising AI tools so they actually produce something usable. For agencies, it is the hidden labor hiding inside every “instant” AI deliverable.
Also known as AI babysitting, model minding, AI supervision overhead
Botsitting is the time a person spends getting an AI tool to do its job properly: feeding it context, checking what it returns, correcting the mistakes, and rerunning prompts until the output is genuinely usable. The model does the typing in seconds. A human still does the supervising, and that supervising is real work that rarely shows up on a timesheet.
The term caught on because the labor is mostly invisible. A draft appears almost instantly, so the work looks finished. The hours that follow, spent reading every line, catching the confident errors, and tightening the inputs, are where most of the effort actually lives. Naming it as botsitting makes that effort visible so teams can plan and price around it instead of pretending it does not exist.
Agencies sold themselves and their clients on AI speed. Botsitting is the line item that speed forgot to mention.
It quietly eats your margin. If a “ten minute” AI task actually costs two hours of review and rework, the savings you scoped into the retainer were never real. Knowing the true cost lets you price and staff honestly.
It changes who you hire. The valuable person is no longer the fastest drafter. It is the one with the judgment to spot a fabricated stat, an off-brand line, or a legal risk before it ships to the client.
It is reducible, not fixed. Better prompts, shared context, and tighter source documents cut botsitting time. Treating it as a measurable cost gives you something to actually improve, rather than a vague sense that AI is “more work than expected.”
A copywriter asks an AI tool to draft thirty product descriptions for a retail client, and thirty drafts come back in under a minute. Then the real day begins. Eleven of them describe features the client does not actually sell, several echo a competitor’s tagline structure, and every one needs the client’s banned-words list applied by hand. The writer spends the afternoon checking each line against the spec sheet, rewriting the invented claims, and rerunning the weak batches with a tighter prompt and the brand guide pasted in. The deliverable that looked finished at 9am is finally clean by 4pm. That afternoon, all of it, is botsitting, and it is exactly the cost a smart agency learns to scope, price, and shrink.
The automations and agents module of the workshop teaches you how to build AI workflows that compress the busywork without taking the craft out of the studio.