Career

“Building products with AI” means AI changes how a team builds products, and the product it ships may contain no AI at all; “building AI products” means AI has entered the process through which users get value and become part of what the product can do.

“Didn’t pass” doesn’t tell me what to do next, so I sort the reasons into three types: Capability gap, Interview gap, and Fit gap.

A few times in interviews I mixed things up: they asked about Prioritization and I was still talking about signal quality and validation; they asked about Discovery and I jumped too fast to value, cost and priority.

After a few rounds of AI PM interviews, I’ve found that the same title can mean very different work: enterprise delivery, product/feature, platform, or exploration.

For a traditional PM moving into AI, don’t start from “what does an AI PM need to know,” a list you’ll never finish. First take stock of your most defensible experience, then pick one AI product layer that amplifies it.

Among 43 AI product roles, I kept seeing once-rare roles like Evals, Harness and Forward-Deployed: problems someone used to handle “on the side” have become complex enough to need a dedicated owner.

After going through 43 AI product JDs, I broke “technical ability” into four layers: AI literacy, builder fluency, system fluency, and ML / infrastructure depth.

After going through 43 public job postings, I stopped sorting by title and sorted by “what this PM is actually responsible for,” ending up with seven kinds: model, platform / harness, evals / safety, applied, vertical, enterprise transformation, and forward-deployed.

The first wave of AI’s impact may not be jobs disappearing, but the deliverables we used to prove our professional worth — PRDs, mockups, code — rapidly getting cheap. The standard is shifting from “what you delivered” to “what results you drove.”