Product Strategy and North-Star Metric for a Shared In-House AI Agent Platform

You are the product manager for an internal AI platform: other teams across the company use it to build their own AI agents. Answer in two parts. ### Clarifying Questions - How far along is the platform: are teams already running agents on it in production, or are most still building their own stacks or using outside tools? - What does the platform provide today: model access only, or also connectors to internal systems, permission handling, evaluation and monitoring? - Who builds on it: engineering teams shipping customer-facing agents, internal-tools teams automating operations, or non-engineers in business functions? - Is using the platform required by leadership, or does it have to win teams over against building their own? ### Part 1 — Strategy What is your strategy for the platform: who it serves first, what it does for them, and how it gets adopted? ```hint Why not build it themselves? Any team can call a model API and wire up its own tools. Think about what the platform can offer that a team cannot easily get alone, and at which step between a working prototype and a trusted production agent that matters most. ``` #### What This Part Should Cover - The platform's purpose for the company and the builder segment it serves first, with the reasoning - The biggest obstacles builders face and which one the platform attacks first - What makes the platform hard to replace, and the plan for adoption ### Part 2 — The one metric If you could own only one metric for the platform, which would it be, and why that one? ```hint Activity is not value API calls, agents created and registered developers can all grow while nothing useful happens. Look for a number that moves only when agents built on the platform do real work well. ``` #### What This Part Should Cover - A precise definition of the metric: what counts and what does not - Why it beats the obvious alternatives - The supporting metrics and guardrails that keep it honest ### What a Strong Answer Covers - A strategy and a metric that follow from each other - The realities of an internal platform: other teams as customers, cost paid centrally or charged back, and competition from building in-house or adopting outside tools - Security, permissions and data access for agents that act on internal systems - Sequencing: what ships first and how the platform earns adoption before broad rollout ### Follow-up Questions - Your metric is climbing, but nearly all the growth comes from one team. Is the strategy working? - A large team wants to use an outside agent framework instead of your platform. How do you respond? - Model spend on the platform doubles in a quarter. How do you tell whether that is good news?