Linear · AI product · Mid · 45 to 60 minutes

How should Linear scale backlog triage

Linear is a software product management and issue tracking tool used by teams to plan and ship work. Answer this product management case and get it graded line by line against a weighted rubric.

Product strategy
AI UX
Prioritization
Metrics
Decision making

The situation

Linear's product leadership asks you to recommend whether and how the company should help teams triage incoming issues with AI. Customers use Linear as a system of record for bugs, requests, and product work, but many backlogs become noisy over time. Similar issues are scattered across projects, labels are applied inconsistently, and a small number of people end up relying on memory to decide whether a new issue is a duplicate, related to existing work, or should be routed elsewhere. That makes the backlog less useful for prioritization and trend spotting. This question has become urgent because Linear is already improving its search infrastructure, which could make historical backlog data easier to retrieve and compare against new issues. At the same time, modern language models make richer suggestions possible, but they can be slower, more expensive, and harder to trust than deterministic product features. Assume larger customer workspaces may receive hundreds of new issues per week. Your recommendation needs to balance usefulness, speed, trust, and fit with Linear's existing workflow, while deciding how much control stays with the human triager at launch.

What to produce

A 1 to 2 page product strategy memo.

Your answer must

Define the primary user, the core triage problem to solve, and why backlog organization matters for downstream product decisions.
Recommend an MVP scope for the AI feature, including which triage decisions to support first and which adjacent capabilities to defer.
Outline a high level approach for how the system would use existing backlog and search data to generate suggestions, with key tradeoffs across quality, latency, and cost.
Describe the user experience principles that would make suggestions trustworthy, including how people review, accept, reject, or understand them.
Specify success metrics, risks, and a phased rollout or experiment plan that would let Linear learn safely before expanding automation.

Constraints

The launch must keep a human in control of final triage decisions.
The feature must fit naturally into Linear's existing triage workflow without adding major screen clutter.
AI generated suggestions must be clearly distinguishable from fields set by humans or existing rules.
The experience must still feel responsive even if AI reasoning takes longer than Linear's usual product interactions.

Format

Use short sections for problem, recommendation, tradeoffs, user experience, and rollout.

How Linear actually played it

Linear explains how it built Triage Intelligence to organize backlogs by grouping similar issues and applying labels consistently, work that used to depend on a few people's institutional knowledge. The post shows the problem framing behind the feature and how the team made AI suggestions useful for product decisions. A good example of an AI feature built around a manual chore.

Pick an AI use case that removes a known manual chore.
A backlog is only a signal if it is organized consistently.
Keep the human in control of final triage decisions.

Built from How we built Triage Intelligence by Yann-Edern Gillet and Matthijs Wolting.