Guides

AI Triage

On the AI Triage plan, every submission arrives already worked: summarized, labeled, prioritized, deduplicated, and packaged for an AI coding agent. You open the ticket to fix it, not to figure out what it is.

01How it works

How it works

  1. 01A customer submits your form. Issued creates the Linear or GitHub issue exactly as on Core — triage never delays or blocks issue creation.
  2. 02The moment the issue exists, the submission is read by the model: summarized, categorized, prioritized, checked against your recent reports, and scanned for missing information.
  3. 03Results are applied to the issue: labels, priority, a triage comment with agent handoff context, and duplicate links.
  4. 04If key details are missing and the customer left an email, a clarifying follow-up is sent automatically — their reply lands back on the issue.

Triage runs asynchronously after the issue is created — results typically land within a minute. If a triage run fails, it retries and your issue is untouched either way; failures never affect issue creation or email replies.

02What lands on the issue

What lands on the issue

  • Labels — suggested from your repo or team's existing labels only, so your taxonomy stays yours.
  • Priority — Linear gets its native Urgent/High/Medium/Low field; GitHub (which has no priority concept) gets a label like priority: high.
  • Duplicate grouping — a native duplicate of or related relation on Linear; cross-linking comments on both issues on GitHub. Nothing is ever closed, merged, or moved automatically.
  • The triage comment — a summary, the suggested metadata, any duplicate note, and the agent handoff context. It's posted by the Issued bot, so it never emails your customer.
03Agent handoff context

Agent handoff context

Every triage comment includes a self-contained markdown brief — what's broken, repro details, environment, customer impact — written for an AI coding agent. On GitHub it's collapsed behind adetailstoggle; on Linear it's a section of the comment. Paste it into Claude Code, Cursor, or Devin, or point the agent at the issue.

agent context (excerpt)
## Export Button Disabled on iPad

**What's broken:**
Export to CSV button is greyed out on iPad (iOS 17.5, Safari),
but works normally on Mac. User is on Pro plan, app v2.1.0.

**Repro:** log in on iPad Safari → open dashboard → Export
**Impact:** 2 related reports this month
04Clarifying emails

Clarifying emails

When a report is missing something you'd need to act (an OS version, repro steps), and the customer left an email address, Issued sends one automated follow-up asking for exactly that. It never asks for anything the customer already provided in the form, and it sends at most one clarifying email per submission — ever.

  • The email is clearly marked as an automated follow-up and threads under the ticket — the customer just hits reply, and the answer lands on the issue like any other reply.
  • Prefer to review first? Turn off Auto-send clarifying replies in project settings — the draft then appears inside the triage comment, and posting it as a regular comment sends it.
05Settings

Settings

Everything is on by default and configurable per project, under the project's edit page. Each action can be switched off independently:

AI Triage Master switch for the whole pipeline on this project.

Apply labels Adds suggested labels — always chosen from your existing label set, never invented.

Set priority Native priority on Linear; a priority label on GitHub.

Post triage comment The summary and agent handoff context on each issue.

Group duplicates Duplicate/related relations on Linear, cross-link comments on GitHub. Issues are never closed or merged automatically.

Auto-send clarifying replies Off = drafts appear inside the triage comment instead of being emailed.

Related

AI Triage builds on the email loop and provider integrations.