AI Product Manager who ships

Fair question, and I hear it weekly. My honest answer: the title is new, the work is not. Every PM I know who ships AI features has, by necessity, learnt to own model choice, prompts, evals and telemetry. Whether their business card says "AI PM" or "Senior PM" is a branding choice.

Where the role is genuinely different: the ratio of time spent on prototyping, evaluation and guardrails. In a traditional PM week I might spend 10% on prototyping. In an AI PM week it's 40%, because I can, and because eval quality is now a product KPI.

So yes, it's real. And no, you don't need to hire an "AI PM" if your existing product team has picked up the tools. What you do need is someone accountable for the AI product surface: its accuracy, its cost per query, and its failure modes. That accountability is the job.