Introduction
Delegating issues to coding agents is now central to how Linear is used. Teams assign issues to Linear Agent through Coding Sessions, or to third-party agents like Cursor, GitHub Copilot, Devin, Codex and Claude through Linear's agent platform, and the agent is expected to open a pull request that a human reviews and merges. Issues with a PR attached grew roughly 7x in 2026, so the share of agent-delegated issues that end in a merged PR is one of the clearest signals that delegation actually works. A drop from 62% to 48% in three weeks is a 14-point fall, large enough that teams will notice wasted runs and may stop delegating.
I'll cover clarifying questions, external factors, the delegation journey, metric breakdown, data gathering, hypotheses, root cause analysis, validation, and a decision framework with a resolution plan.
Step 1
Clarifying Questions (3 minutes)
Why it matters: a fixed window by cohort is stable; an open-ended window is biased against recent weeks because PRs have not had time to merge. Expected answer: merged within 14 days, grouped by delegation date. Impact on approach: I'd check whether the last two weeks are fully matured before trusting the drop.
Why it matters: one agent points to that agent's release or integration; all agents point to Linear or GitHub. Expected answer: all agents down, some more than others. Impact on approach: I'd look for shared causes first.
Why it matters: a mix shift (more, easier-to-fail issues) can lower the rate without any agent getting worse. Expected answer: delegated volume grew about 35% over the same period. Impact on approach: mix shift becomes a lead hypothesis.
Why it matters: these are three very different problems: quality, agent failure, and tracking. Expected answer: unknown; needs a breakdown. Impact on approach: the funnel split comes first.
Practice similar questions
Subscribe to access the full answer
.png){kind=logo}
