Jira Service Management Integration

What TR-AI does on a request

Step 1 — Prepare an agent account and API token

1. 1
   Pick the account

   Use an Atlassian account with an agent licence on the service project (a dedicated account like tr-ai@yourcompany.com makes TR-AI's comments easy to spot).
2. 2
   Create an API token

   Signed in as that account, go to id.atlassian.com → Security → Create and manage API tokens and create a token.
3. 3
   Find the service desk ID

   Open the customer portal; the number in /servicedesk/customer/portal/<id> is the service desk ID.

Step 2 — Enter the credentials in TR-AI

Choose Jira Service Management at sign-up (or Update credentials on the dashboard's setup card) and enter the site URL (https://acme.atlassian.net), the account email, the API token and the service desk ID. TR-AI checks the token and that the account can see the service desk before saving.

Step 3 — Add the two automation rules

In the service project open Project settings → Automation → Create rule. Your dashboard shows the exact URL (with its secret key) to use in both rules.

1. 1
   Rule 1 — new requests

   Trigger Work item created. Action Send web request: URL = your TR-AI web request URL, method POST, header Content-Type: application/json, Web request body = Custom data with the body below. Name it, turn it on.
2. 2
   Rule 2 — comments

   Trigger Work item commented. Optionally add the condition Comment visibility = public. Same Send web request action with the second body below.
3. 3
   Rule actor

   Leave the rule actor as Automation for Jira. TR-AI's own comments reach Rule 2 too — TR-AI recognises them by author and ignores them.

{
  "event": "created",
  "issueKey": "{{issue.key}}",
  "issueId": "{{issue.id}}",
  "summary": "{{issue.summary.jsonEncode}}",
  "description": "{{issue.description.jsonEncode}}",
  "reporterName": "{{issue.reporter.displayName.jsonEncode}}",
  "reporterEmail": "{{issue.reporter.emailAddress}}",
  "reporterAccountId": "{{issue.reporter.accountId}}"
}

{
  "event": "comment",
  "issueKey": "{{issue.key}}",
  "issueId": "{{issue.id}}",
  "summary": "{{issue.summary.jsonEncode}}",
  "description": "{{issue.description.jsonEncode}}",
  "reporterName": "{{issue.reporter.displayName.jsonEncode}}",
  "reporterEmail": "{{issue.reporter.emailAddress}}",
  "reporterAccountId": "{{issue.reporter.accountId}}",
  "commentId": "{{comment.id}}",
  "commentBody": "{{comment.body.jsonEncode}}",
  "commentAuthorAccountId": "{{comment.author.accountId}}",
  "commentInternal": "{{comment.internal}}"
}

Step 4 — Create the triage queue

Requests TR-AI can't answer get the label tr-ai-triage. In Queues → New queue create "TR-AI triage" with JQL labels = tr-ai-triage AND statusCategory != Done.

Step 5 — Import history and go live

• TR-AI imports resolved requests from the selected window. The answer is the last public comment by an agent; internal comments are stored as internal context only and never quoted to customers.
• New accounts start in assisted mode: drafts appear as internal comments and in the dashboard for approval.
• Approving a draft posts it as a public comment and transitions the request; rejecting adds the triage label.

How the resolve transition is chosen

Workflows differ, so TR-AI reads GET /rest/api/3/issue/<key>/transitions and picks a transition into a status in the Done category, preferring names containing "Resolve", then "Done", then "Close" — never Cancel, Decline or Won't do. If the transition screen requires a resolution, it uses Done / Fixed / Resolved. If no such transition exists from the current status, the comment is still posted and the request is left in its status.

Troubleshooting