Using AI in Your PM Workflow and Judging Where AI Belongs in a Product

How do you use AI, both in your own work as a product manager and in the way you think about building products? ```hint Concrete tasks, then limits Anchor on two or three specific tasks in your week where AI changed how you work, and for each be ready to say where you stopped trusting its output and why. ``` ### Clarifying Questions - Should I cover personal productivity, building AI features into products, or both? - Should the product examples fit an observability platform, where users are engineers investigating incidents, or can they come from any product I have worked on? - Are there policies I should assume about putting customer data or unreleased plans into third-party AI tools? ### What a Strong Answer Covers - Specific, hands-on uses of AI across the PM workflow (discovery, analysis, writing, prototyping), with what changed in speed or quality. - Verification habits and the limits encountered, such as invented facts, missing context and the confidentiality of customer data. - A point of view on when a product problem calls for AI and when a deterministic rule or a simpler feature is the better choice. - How AI features are evaluated and measured once built: quality bars, user trust, latency and cost. - Relevance to a platform whose users rely on accurate answers while an incident is in progress. ### Follow-up Questions - Tell me about a time AI output led you toward a wrong conclusion. How did you catch it? - An engineer on your team calls a proposed AI incident-summary feature hype. How do you decide whether to build it? - How would you measure whether an AI assistant inside a monitoring product is actually helping its users?