Stale Jira tickets sorted, answered or closed by AI

Stale Jira tickets sorted, answered or closed by AI

An n8n workflow that runs once a day and picks up Jira tickets that have sat unresolved for a week. For each ticket it reads
the comments and asks a model which of three states the ticket is in. A resolved ticket gets a sentiment check, then a
closing comment or a Slack alert. A ticket waiting on the customer gets a reminder. A ticket nobody has answered goes to an
AI agent that searches similar tickets and Notion, posts what it finds and closes the ticket.

What is inside

• workflow.json: the workflow as n8n exports it, 29 nodes and seven sticky notes. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• Both Slack nodes had the author's own channel selected. The channel is cleared, so pick yours.
• Every Jira, Slack, Notion and OpenAI node lost its link to a credential in the author's n8n. Select your own.
• Nothing else. The file has no Code node, and its expressions needed no repair.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• A Jira Software Cloud site and an API token for a Jira credential.
• An OpenAI API key that can use gpt-4o-mini.
• A Notion integration, shared with the pages the agent may search.
• A Slack workspace, an app that can post messages and a channel for the alerts.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add a Jira credential and select it in all nine Jira nodes: Get List of Unresolved Long Lived Issues,
   Get Issue Comments, Send Reminder, Add Autoclose Message, Ask For Feedback Message, Reply to Issue,
   Close Issue, Close Issue2 and the tool Find Simlar Issues.
3. Add an OpenAI credential and select it in OpenAI Chat Model, OpenAI Chat Model1, OpenAI Chat Model3 and
   OpenAI Chat Model4.
4. Add a Notion credential and select it in Query KnowledgeBase.
5. Add a Slack credential and select it in Notify Slack Channel and Report Unhappy Resolution. Choose your channel in both.
6. Open Close Issue and Close Issue2. They move a ticket to transition id 31, named Done in the author's project. Pick
   your own Done status in each.
7. In Ask For Feedback Message, replace link/to/review_site with the address of your review page.
8. In Get List of Unresolved Long Lived Issues, narrow the JQL to a test project while you try it. Save the workflow, which
   calls itself by id, then run it from the canvas.
9. Once the first tickets look right, set the schedule you want in Schedule Trigger and publish.

How a run goes

1. Schedule Trigger fires. Its interval has no settings in the file, so it uses the node's default: once a day at midnight.
2. Get List of Unresolved Long Lived Issues asks Jira for tickets in To Do or In Progress created at least seven days ago,
   at most ten.
3. Execute Workflow starts this same workflow once for each ticket, at Execute Workflow Trigger.
4. Get Issue Metadata keeps the key, title, link, date, reporter and description. Get Issue Comments fetches the comments,
   Join Comments gathers them and Simplify Thread For AI turns them into a short thread and a topic.
5. Classify Current Issue State (Text Classifier) sorts the ticket as resolved, pending more information or
   still waiting. A ticket the model cannot place is dropped, which is the node's default fallback.
6. Resolved: Customer Satisfaction Agent (Sentiment Analysis) reads the thread. Positive goes to Ask For Feedback Message
   and Close Issue. Neutral goes to Add Autoclose Message and Close Issue. Negative goes to Report Unhappy Resolution,
   a Slack alert, and the ticket stays open.
7. Pending: Last Message is Not Bot lets the ticket through only when the last comment does not contain the automated
   note. Issue Reminder Agent writes a short reminder and Send Reminder posts it.
8. Still waiting: KnowledgeBase Agent searches similar tickets with Find Simlar Issues and pages with Query KnowledgeBase,
   and answers in a set shape: solution_found, short_summary_of_issue and response. When it found a solution,
   Reply to Issue posts it and Close Issue2 closes the ticket. When it did not, Notify Slack Channel alerts the channel
   first, then the same two nodes post the agent's answer and close the ticket.

Change the look

• Which tickets: the JQL and the limit in Get List of Unresolved Long Lived Issues.
• When it runs: Schedule Trigger.
• The states: the categories and their descriptions in Classify Current Issue State.
• The wording of every comment: the text in the Jira comment nodes. Each ends with a note that it is automated.
• The agent's rules: the system message in KnowledgeBase Agent.
• Where it looks: the JQL in Find Simlar Issues and the search in Query KnowledgeBase.
• The models: the four OpenAI Chat Model nodes.

Good to know

• It writes to real tickets. Comments are posted and tickets are closed with no one reading them first, so try it on a test
  project and read each comment before you widen the JQL.
• OpenAI's deprecations page, read on 12 November 2026, gives 14 January 2027 as the shutdown of gpt-4o-mini-2024-07-18 and
  names gpt-6-luna as the replacement. The four model nodes still call gpt-4o-mini, so change them before that date.
• The transition id 31 is the author's. A project whose Done has another id needs its own pick in both Close nodes.
• A ticket with no comments is the still waiting case, but an empty comment list gives the next node nothing to run on and
  n8n then ends the branch. Try that case with a real ticket before you rely on it.
• Execute Workflow calls this workflow by its own id, so a copy of the workflow works only once it is saved.
• The review request is posted to every ticket with positive sentiment, and its link starts as a placeholder.
• Each ticket costs one classification call plus a sentiment, reminder or agent step. The agent can make several calls and
  tool searches.
• Ticket text goes to OpenAI, ticket summaries go to Slack, and the agent's searches reach your Jira and Notion.