How do you handle disagreement with a product manager regarding the scope or safety guardrails of an AI feature?
