North-Star and Guardrail Metrics for Shared Team Workspaces in ChatGPT

Suppose OpenAI adds collaborative workspaces to ChatGPT, so that a team can work with the assistant in a shared space instead of each person working in a private chat. You are the product manager for this launch. Explain how you would define success for the feature and how you would measure it. ```hint Find what only a team can get Ask what a group gets from a shared workspace that each member could not get alone from a private chat, then look for a measure that moves only when that shared value is happening, not when one person simply uses ChatGPT more. ``` ### Clarifying Questions - What does a workspace contain at launch: shared conversations, shared files and instructions that ground the assistant's answers, live co-editing, or a mix? - Which teams is it for: organizations already paying for business seats, small groups of individual subscribers, or both, and is it a paid add-on or included in existing plans? - Over what horizon is success judged: early adoption in the first weeks, or retention and revenue a few quarters after launch? - Which privacy or admin commitments to business customers limit what we can log or inspect when measuring usage? ### What a Strong Answer Covers - A stated goal for the launch, tied to why OpenAI would want teams rather than only individuals using ChatGPT, before any metric is chosen - A north-star metric defined at the team or workspace level that rewards repeated shared use, with a precise definition rather than raw message volume - A funnel of supporting metrics from workspace creation through teammate activation to habitual use, plus lagging business outcomes such as seat growth and plan retention - Guardrails for cannibalized individual use, answer quality, permission or data-exposure incidents, and serving cost - A measurement plan: baselines, how to estimate the causal effect when already-engaged teams adopt first, and the results that would change the roadmap ### Follow-up Questions - A month after launch, most workspaces have only one active member. What do you conclude, and what do you investigate first? - How would you show that workspaces improve retention of paying teams, rather than simply being adopted by teams that were going to stay anyway? - If your north-star metric rises while messages per user fall, is that good or bad news, and how would you tell?