12. AI Product Manager
Evidence label: Role family. AI Product Manager and product-manager titles scoped to AI are current, but compensation premiums require employer- and level-specific evidence.
What they do
Product management with deep AI fluency. They understand what AI can and can't do, manage the unique product challenges of AI systems (non-determinism, latency, hallucinations, cost), and translate between business needs and AI capabilities.
Responsibilities
- Define product requirements for AI features with technical constraints in mind
- Work with AI engineers on evaluation criteria (what does "good" look like?)
- Manage the unique UX challenges of AI: uncertainty, latency, error handling
- Own the cost/quality/speed tradeoffs for AI features
- Communicate AI limitations and risks to stakeholders
- Run A/B tests on model versions, prompt changes, feature changes
What makes AI PM different from traditional PM
Traditional PM ships features that behave deterministically. AI PMs ship systems where outputs vary. They need to think probabilistically: not "will this work?" but "what % of the time will this work, and what happens in the other cases?" Quality measurement is continuous, not binary.
Required skills
Standard PM skills (roadmapping, prioritization, user research) plus: LLM API familiarity, eval design, basic Python for running experiments, understanding of model tradeoffs (accuracy vs. cost vs. latency), AI UX patterns.