Jargon:Meta-work
1 min read
Say this at the party
“The work of managing AI rather than doing the task yourself: setting it up, checking what it made and steering it.”
In plain English
When an AI does the writing, coding or research, your job changes. You write the brief, answer its questions, check the result and decide what's next.
That layer of managing is meta-work. It sounds like admin, but it's where the judgement now sits.
Do it badly and you become the bottleneck: lots of agents, all waiting on you.
Where it came from
The word has been around in workplace research for years. Anthropic's economists gave it new life in October 2026, in a study of their own staff: meta-work rose from 10% to 22% of people's hands-on time with agents in a year, and agents spent about 80% of their working time waiting for a person to reply.
See it in action
You have three agents on the go: one drafting a report, one fixing the website, one sorting invoices. You spend the morning answering their questions and checking their work. That's meta-work.
What people get wrong
- Meta-work isn't wasted time. Checking and steering is what stops fast AI from making fast mistakes.
- More agents doesn't mean less work for you. Past a point, each new one adds more checking than it saves.
Where Ugenie fits
Meta-work is what Ugenie is for. It keeps an eye on every AI you use and brings you the three main tasks that need you next.
Get Ugenie →Go deeper
- PaperAnthropic Economics team, 2026How employment is changing at AnthropicThe numbers: six hours of agent work for every human hour.
- TalkMaggie Appleton, 2026One Developer, Two Dozen Agents, Zero AlignmentThe team version of the problem.
- ArticleEthan Mollick, One Useful Thing, 2026A Guide to Which AI to Use in the Agentic EraHow to get good at managing AI, not just prompting it.
Where it comes up
- Your new job is managing the AIsAnthropic studied its own staff. Agents now do six hours of work for every human hour, and people spend more of their time on “meta-work”.Our take →
- One developer, two dozen agents, zero alignmentAI agents are built for one person working alone. Real work is done in teams, and that's the new bottleneck.Our take →
- Stop prompting, start managingEthan Mollick's guide to picking an AI now that they do whole tasks: think about the model, the app and the harness.Our take →
Related jargon
Drafted with AI, checked by the AI Council (Claude, ChatGPT and Gemini) and signed off by Phin.