Redesign Netflix Member Support With AI Assistance

Netflix wants to use AI to make getting help easier for its members. How would you improve the Help experience? ```hint Start where the problem happens Picture where a member is when something breaks, often in front of a TV with only a remote, and list what Netflix already knows at that moment about the device, the error and the account. ``` ### Clarifying Questions - Which kinds of help are in scope: playback and device troubleshooting, account and billing questions, contact with human agents, or all of them? Is support for creators or partners out of scope? - How does Help work today: a searchable help center of articles, with chat or phone agents for harder cases? - What should improve first: how often members fix problems on their own, how fast they get a fix, satisfaction, or support cost? - May the AI take actions on a member's account, such as signing out devices or updating a payment method, or only explain and guide? ### What a Strong Answer Covers - A clear scope for "Help" and a goal that measures resolution, not just shorter help sessions. - A chosen member segment, such as less tech-savvy members on TVs, and the pain points along their path from problem to fix. - AI solutions that use device, error and account context, with a safe handoff to human agents, prioritized by impact and effort. - A resolution-based north star with guardrails for wrong answers, repeat contacts and cancellations. - Risks specific to AI support: confident wrong answers, account takeover through the assistant, privacy, and making cancellation harder. ### Follow-up Questions - After launch, contacts to human agents fall, but members who used the assistant cancel at a higher rate. What do you investigate and change? - How would you evaluate the assistant before it reaches members, and how would you keep evaluating it after launch? - Should the assistant be allowed to process a cancellation or a plan downgrade on its own? What safeguards would you require?