What we can and cannot tell you
No company publishes its AI PM loop in full, and loops differ by team and level. Below are the conversations that recur in public candidate reports and preparation material, with what each one tests. Confirm your own schedule with your recruiter, and ask whether AI tools are allowed in any build exercise.
Recruiter and hiring manager screens
Expect to walk through one AI product you worked on: what the model did, how you knew it was good enough, and what it cost to run. If you have not shipped one, a prototype or an evaluation you ran will do, provided you have numbers.
Product sense
The classic prompt with an AI layer. Design an assistant for a group of users, or decide whether a feature should use AI at all. Strong answers choose a user and a painful job, separate the parts a model should handle from the parts a rule handles more safely, set a quality bar and design what happens when the model is wrong.
Evals and launch decisions
Increasingly a conversation of its own. You may get a feature and be asked how you would know it is ready, or see a results table and be asked whether to ship. Expect follow-ups on sample size, how the judge was validated and which segment is weakest.
Metrics and unit economics
Define success for an AI feature, then defend its economics. The usual trap is a metric that rewards using the AI rather than the outcome users came for, such as counting summaries generated instead of tasks finished.
Technical and system design
Not a coding interview. You design an AI system in components and trade-offs: retrieval, model routing, caching, latency budgets, fallbacks and permissions. What is graded is whether you can tie each choice to quality, cost or risk.
Live prototyping
If the recruiter confirms a build exercise, establish the duration, permitted tools and expected deliverable. Practise a narrow workflow, test deliberate inputs and explain what you would change. Our timed practice plan is a suggested rehearsal, not an employer assessment specification.
Behavioural and values
Prepare true stories about launches you stopped, disagreements, ethical calls and times evidence changed your view. Read the employer's published principles and connect them to decisions you have made. Confirm whether these themes have a separate conversation in your interview schedule.
Splitting your preparation
The chart below is a starting split of practice hours. It is not how any employer weights its loop. Move hours toward whatever your first mock answers expose.
Sources and further reading
Questions about this topic
It varies by company, team and level, and we do not publish counts we cannot verify. Ask your recruiter for the schedule and the format of each conversation.