Linear · AI product · Senior · 45 to 60 minutes

How should Linear scope an AI agent?

Linear makes software for teams to plan, track, and manage product development work. Answer this product management case and get it graded line by line against a weighted rubric.

Product strategy
AI product judgment
Risk management
Prioritization
Systems thinking

The situation

You are a senior PM at Linear working with the founders and engineering leads on a potential AI agent inside the product. The team believes AI could help users get meaningful work done across planning and issue tracking, not just answer questions or generate text. The open question is what kind of product to build first. One camp wants a tightly defined assistant with a small set of scripted workflows so behavior is consistent and easy to explain. Another camp wants a more capable agent that can handle a wider range of user intents and edge cases, even if outcomes are less predictable. The stakes are high because Linear is trusted for core workflow data and actions, and a poor first release could hurt user trust or make the product feel gimmicky. At the same time, a product that is too narrow may miss the real upside of AI and fail to stand out. The founders ask you for a recommendation on product direction, scope, and rollout approach. They want a proposal that balances user value, trust, and the realities of current model behavior.

What to produce

A 2 page product strategy memo.

Your answer must

Define the user problems this AI product should solve first, and name the primary user or team context you would target.
Compare at least two plausible product approaches for the first release, including a tighter workflow driven assistant and a more open ended agent, then recommend one.
Specify the operating boundaries of the product you recommend, including what it can do, what actions need confirmation, and what it should never do in v1.
Propose a rollout and sequencing plan for v1 to v2, including how you would learn safely before expanding capabilities.
Define success metrics that cover both value and trust, and include the main risks and how you would monitor them.

Constraints

The product must fit into Linear's existing core workflows and cannot depend on a separate human review step for every action.
You must preserve user trust around real work data and actions, so mistakes have higher cost than in a toy assistant.
Model behavior is improving quickly but remains unreliable in edge cases, so your plan cannot assume perfect consistency.

Format

Use clear headings and concise bullets. A small comparison table for the approach options is welcome.

How Linear actually played it

Linear describes building an agent by setting boundaries through its system prompt, tools and guardrails, rather than scripting fixed paths, because the value lies in work the team did not anticipate. The post shows the product trade-off between consistency and flexibility. Good for PMs scoping AI agents.

Define boundaries for an agent rather than a rigid path
AI flips the usual goal of making outcomes predictable
Tool design is a product decision as much as an engineering one

Built from How we built Linear Agent by Matthijs Wolting.