Uber Product Design: Build an AI-Powered Product for Riders or Drivers

You are interviewing for a product manager role at Uber. Design an AI-powered product for a ride-sharing app: choose who it serves and the problem it solves, describe the experience, and explain how you would judge whether it works. ```hint Start from a marketplace problem Pick one group on either side of the ride-sharing marketplace and one frustrating moment in their journey before thinking about models, then check whether a simpler, non-AI fix would solve it just as well. ``` ### Clarifying Questions - Should the product live inside the existing rider or driver app, or can it be a separate product? - Is the main goal more completed trips, lower cost to serve, or greater trust and safety? - Which side of the marketplace is in scope: riders, drivers, or both? - Are there limits on what drivers may interact with while driving that the design must respect? ### What a Strong Answer Covers - A chosen user segment and pain point, justified by its effect on the whole marketplace - Why AI beats a simpler rule or interface change for this problem, and which models or capabilities it needs - A concrete product experience, including what the first version includes and leaves out - Marketplace side effects, since a change for one side affects the other, and how to test for them - A primary metric with guardrails, and the main risks: safety, privacy, fairness and wrong AI output ### Follow-up Questions - Users in the treatment group of a ride-sharing experiment compete with control users for the same trips. How would you design the experiment? - How would you show that the AI adds value beyond a simple rule-based version? - What would you do if the product improved outcomes for its users but made them worse for everyone else?