Scharlette, love this. Reading the codebase to learn the product in two days is a real edge, and turning it into docs the team didn’t have is the kind of thing I’d put on a resume. On the title question, I think “AI PM” covers two different jobs that people mash together: 1. PMs who build AI products. They own model behavior, evals, failure modes, cost and latency tradeoffs, and how users come to trust the output. 2. PMs who use AI to do the PM job. That’s what you just described, and it’s quickly becoming table stakes for every PM. Most job posts and interview loops that say “AI PM” mean #1. So here’s where I would push. You said only one part of your product is agentic. That part is your way in. You don’t need to pick models to own it. Start asking: - How do we know the agent did the right thing? What’s our eval? - What happens when it’s wrong, and who notices? - What does each run cost, and is it worth it? - Where does the user need to stay in the loop? Own those answers for even one feature and you’ll have a real AI PM story to tell, not just the title. One caution on the code-generated docs: code tells you what the product does, not why. Run them past an engineer before they become the source of truth. Otherwise you’ll end up writing down accidents as if they were decisions.