Picking and Designing the First AI Agent Use Case for a Streaming Platform

A streaming service wants to build an AI agent. You are the product manager: decide what the agent should do, and design it. ```hint Choose the job before the features An agent could serve viewers, subscribers with account problems, or internal teams. Before designing capabilities, pick one job and explain why acting on someone's behalf beats the menus and search the service already has. ``` ### Clarifying Questions - Is this a video, music or live TV service, and is it a paid subscription, ad-supported, or both? - Is the agent meant for subscribers, for example to find something to watch or fix an account problem, or for internal teams such as customer support? - What may the agent do on its own: only inform and recommend, or also change plans, billing and devices? - Which surfaces must it work on: TV apps driven by a remote, mobile, web, or voice? ### What a Strong Answer Covers - A justified choice of job and user segment, tied to what drives a streaming business, such as retention, engagement and support cost - The agent's capabilities and limits: which actions it may take, what needs confirmation, and when it hands off to a person - An experience shaped by streaming surfaces and the user's moment, such as a remote control on a TV versus a phone - Success metrics with guardrails, including retention effects and the rate of wrong or unwanted actions - Agent-specific risks: wrong billing or account changes, false claims about titles, privacy of viewing history, and pressure tactics around cancellation ### Follow-up Questions - A subscriber asks the agent to cancel. Should it try to keep them, and how would you stop that from feeling like a trap? - How would you test the agent before launch, given that it can change real accounts? - How does the design change for a household account shared by several profiles, including children's profiles?