Should you tell people you use AI for most of your work?

A PM admitted something a lot of people feel and few say: they use AI for maybe 90% of what they produce, feel like they're cheating, and have caught themselves quickly switching tabs when a colleague approaches so nobody sees the chatbot. Am I the only one, they asked. They are very much not.

The hiding is the wrong response, but the discomfort underneath it is worth taking seriously, because it's pointing at a real question that has nothing to do with shame.

Why the shame is misplaced

Nobody hides that they use a spreadsheet, a search engine, or a calculator. Using a powerful tool to produce work faster has never been cheating; it's the entire history of knowledge work. If using AI to draft a doc is cheating, so was using a word processor instead of a typewriter. The tab-switching is a status anxiety, not an ethics problem — the fear that if people see the tool, they'll think the tool is doing the job and you're just the typist.

The real question the discomfort is pointing at

Here's the uncomfortable part the shame is circling: if AI produced 90% of your output and that output is good, what exactly was your contribution? That's not an accusation — it's the most important question you can ask about your own role, and the people quietly hiding their AI use are avoiding it instead of answering it.

The answer, when the work is actually good, is that your contribution was the part that isn't the typing:

• The judgment about what to make and whether it's worth making.
• The real context you fed the model — the transcript, the constraint, the strategic why — without which it would have produced confident slop.
• The review that caught where it was wrong, because it's always wrong somewhere, and knowing where takes the expertise the model doesn't have.

If that's where your value sits, you have nothing to hide and a clear story to tell: "I use AI to produce drafts fast, and I spend my time on the decisions and the grounding and the review that make the drafts actually right." That's not a confession. It's a description of a more leveraged PM.

When the discomfort is a real signal

And if, honestly, the AI did 90% and the judgment and the context and the review — if you can't name your contribution — then the discomfort isn't shame, it's information. Not that AI is taking your job, but that you've been spending your time in the layer the model is good at instead of the layer it can't reach. The fix isn't to hide better. It's to move up the stack to the work that's actually yours.