AI Product Manager
Spec the outcome. Then let engineering build the system.
For product leaders whose PMs are shipping AI features they cannot specify. This seat owns the task, the eval, the escalation path, and the kill criteria.
My PMs can write a PRD. They cannot tell me when the model is allowed to be wrong.
What the AI Product Manager owns
The outcome
The problem, and the metric that would make the feature worth keeping.
Good enough
The eval, the rollout, the human escalation, and the rule that turns the feature off.
What to build next
Ranked by P&L effect, not by how novel the demo looks.
This seat often opens an engagement, then hands a scoped build to engineering. The person who builds the feature is the Product AI Engineer. A feature PM writing tickets for a chatbot is a different hire.
Where this seat sits
AI Pivot Squad
Primary home. This seat opens the engagement. The squad can build what it scopes.
Product AI Engineer
The person who ships the feature this seat specifies.
Business Process AI
When the outcome is a workflow rather than a product feature.
Common questions
No. This seat defines good enough before anyone writes a prompt. The Product AI Engineer builds the feature. A squad builds it when the scope is larger than one person.
No. A feature PM can write a PRD. This seat can say when the model is allowed to be wrong, and when the feature comes off.
The milestone ends when the outcome, the eval, and the kill criteria are in writing and a builder can start. Staying on to watch the rollout is a later milestone, scoped then.
Define good enough before the build.
A 30-minute scoping call is enough to tell whether this seat is the entry, or whether a builder should come with it.