Docs

AI Triage Agent for Jira

Setup and usage guide

Overview

The AI Triage Agent automatically generates draft replies for incoming Jira Service Management tickets. When a new issue is created, the agent searches your historical resolved tickets and Confluence documentation for relevant context, then produces a cited draft response that your support agents can review and use.

The agent runs entirely within your Atlassian environment using Atlassian Forge. Your Jira and Confluence data is processed by the agent and never shared with third parties except Anthropic's Claude API for draft generation (see Privacy Policy).

What the agent does: new issue created → semantic search over past tickets + Confluence docs → Claude generates a cited draft → draft appears in the issue sidebar for your agent to review.

Requirements

  • Atlassian Cloud (Jira and/or Confluence)
  • Jira Service Management project (the issue panel appears on all Jira issue types)
  • Project Administrator access to configure the app

Installation

1
Go to the Atlassian Marketplace and search for AI Triage Agent for Jira, or follow the direct listing link.
2
Click Get it now and select your Atlassian Cloud site.
3
Review the requested permissions and click Accept. The app requires read access to Jira and Confluence content, and write access to post AI-generated comments on issues.
4
The app is now installed. Navigate to a Jira project to complete configuration.

Configuration

Before the agent can generate drafts, you need to sync your Jira ticket history and Confluence documentation into the agent's knowledge base.

Access the settings page

  1. Open any Jira project.
  2. Go to Project Settings (bottom of the left sidebar).
  3. Under Apps, click AI Triage settings.

Sync your knowledge base

The settings page shows the current sync status and two sync buttons:

  • Sync Jira — indexes resolved Jira tickets from the past 12 months. These are used as reference examples when generating draft replies.
  • Sync Confluence — indexes pages from your Confluence space. These are used as source documentation for answers.

Run both syncs on first setup. Each sync typically completes in 10–30 seconds depending on the volume of content. The page shows the document count and last sync timestamp after completion.

Tip: Run the initial sync from the settings page. After that, the agent syncs automatically every day to pick up new tickets and updated documentation.

How triage works

Automatic draft generation

When a new Jira issue is created in any project where the app is installed, the agent automatically:

  1. Converts the issue title and description into a semantic embedding.
  2. Searches the knowledge base for the 3 most similar resolved tickets and 2 most relevant Confluence pages.
  3. Sends the issue and retrieved context to Claude to generate a draft reply with numbered citations.
  4. Stores the draft and displays it in the issue sidebar.

Viewing the draft

Open any Jira issue. In the right-hand sidebar, look for the AI Triage panel. It shows:

  • The generated draft reply with [1][2] citations linking to source tickets or Confluence pages.
  • Suggested labels, priority, and assignee based on similar past tickets.
  • The source documents used to generate the draft.

Using the draft

Click Use draft to mark it as accepted. Your agent can copy the draft text directly into a Jira comment or reply to the customer. The draft is a suggestion — agents should review it before sending.

Note: If no sufficiently similar past tickets or documentation are found, the agent falls back to a safe escalation message rather than generating a potentially incorrect answer.

Scheduled sync

After the initial manual sync, the agent automatically runs an incremental sync once per day. This picks up Jira tickets and Confluence pages modified in the past 24 hours, keeping the knowledge base current without any manual action.

The last sync timestamp is visible on the settings page. If you add significant new documentation or resolve a large batch of tickets, you can trigger a manual sync at any time from the settings page.

Permissions

The app requests the following Atlassian permissions at install time:

  • read:jira-work — read Jira issues and comments for the knowledge base sync.
  • write:jira-work — post the AI-generated draft as a comment on new issues.
  • read:jira-user — read user display names for suggested assignee.
  • read:confluence-content.summary, read:page:confluence, read:space:confluence — read Confluence pages for the knowledge base sync.
  • storage:app — store the installation identifier within Forge's secure storage.

No Atlassian credentials (passwords or API tokens) are collected or stored by the app. Authentication is handled entirely by Atlassian Forge's built-in OAuth.

Troubleshooting

The AI Triage panel shows no draft

The agent only generates a draft for issues created after the initial sync. Issues that existed before installation will not have a draft. Create a new test issue to trigger draft generation.

Draft says "no relevant sources found"

The knowledge base may be empty or not yet synced. Go to Project Settings → AI Triage settings and run Sync Jira and Sync Confluence.

Sync returns 0 docs

If your Jira tickets or Confluence pages were last updated more than 12 months ago, increase coverage by running the sync after making a small edit to any page or ticket to bring it within the sync window.

Settings page shows "Cannot reach backend"

This indicates a temporary connectivity issue. Wait a few seconds and refresh the page. If the problem persists, contact support.

Support

For questions, bug reports, or feature requests, contact us at support@a2agateway.com.

For security vulnerabilities, please follow our responsible disclosure process outlined in the Security Policy.