top of page
kordic-logo
Contact

Auto-Clean Jira Backlogs with Atlassian Rovo Agents

  • Writer: Jessica Piikkila
    Jessica Piikkila
  • Jun 23
  • 5 min read

Date Tested: June 23, 2026


Is your Jira backlog a growing todo list packed with dead tickets, legacy bugs from last year, and half-baked ideas from old business initiatives? Does your team board load super slow and your saved filters making every other page load slow too?


If the answer is yes then you might find this how-to build an agentic workflow to automatically tidy up your backlog valuable.


Instead of spending additional time or mental energy over manual triage, you can build a machine to do it. By combining Atlassian Rovo Agents with simple Jira Automation, you can spin up an automated vacuum cleaner for your backlog.


This how-to article will show you how to build an agent that can:

  • sweep up stale items

  • flag duplicate bugs

  • check everything against your code standards

  • and hand your team a clean list of recommendations to sign off on in minutes.


Audience:

  • Product Owners & Scrum Masters who want to improve product backlog maintenance using Rovo

  • Jira Administrators that support automated workflows and verified Rovo Agents for end-users

Note to reader :

I used Jira Premium in writing this how-to guide. My version of Jira Cloud may have labels or steps that are different from what you may see on your version or tier of Jira Cloud. Use Rovo to ask for guidance when the terms appear different. For example, 'rules' is called 'flows' in certain versions of Jira Cloud.


Prerequisites

  • Atlassian Cloud Premium or Enterprise with Rovo switched on.

  • Jira Global or Product Admin rights to build global automation sequences.

  • Confluence Space Admin rights to build and store your policies.

Process Workflow & Architecture

Before we configure the rules, look at how the data flows. The automations run in the background, keeping human review at the end of the line:


Step-by-Step Implementation Guide


Step 1: Document Backlog Refinement Rules in Confluence

AI cannot read your mind. If your human engineers cannot agree on what makes a story "ready," a robot won't rescue you. Come up with your raw team working agreements:


  1. Create a Confluence page named Engineering Triage & Backlog Policy.

    1. Optional TIP: Use Rovo generate this page for you. It will create a beautiful page structure with clean headings.


  2. Map out these three key sections:

    • Definition of Ready (DoR): Every ticket needs a clear "As a... I want... So that..." structure. It must have defined Acceptance Criteria, and a linked Epic. No exceptions.

    • Stale Ticket Rule: Any issue in the "Backlog" or "To Do" status that has not been touched, commented on, or modified in 90 days.

    • Duplicate Management: Mark matching issues as duplicates, link them back to the active ticket, and close them down automatically.

  3. Publish the page somewhere visible to your team.

Step 2: Build the Rovo Backlog Agent

The agent acts as your digital back-office manager, comparing active Jira issues against your policy page.

  1. Go to Rovo in your Atlassian menu (or navigate to Apps > Rovo Agents).

  2. Click Rovo dropdown, and select '+ create agent' close the popup box that automatically opens to guide you through crating an agent. Click the blue create button, click on 'Manually create agent'. You will be directed to define all the agent information.

  3. Click and give Agent title "Backlog Hygiene Specialist for [Team Name]"

  4. Click on add a description and paste "Automated assistant that deep-cleans Jira backlogs to protect team focus."

  5. Paste this concrete prompt block into execution instructions [include more details if desired]:


You are a seasoned Agile Scrum Master and Technical Product Owner assistant. Your task is to audit and refine Jira work items to make them actionable for team decision making.

For every issue analyzed, perform these checks:
1. COMPLIANCE CHECK: Compare the issue's Description and Epic Link against our Confluence Engineering Triage & Backlog Policy. Identify missing elements (e.g., missing Acceptance Criteria, story points).
2. DUPLICATE CHECK: Search contextually for similar issues within the same project.
3. FRESHNESS CHECK: Evaluate if the ticket has been idle for >90 days without updates.

Your output response must modify the ticket as follows:
- Add a structured comment starting with "🤖 **Rovo Triage Evaluation**" highlighting gaps.
- If the ticket is stale, recommend moving it to the "Won't Do / Archived" state and label it "stale-candidate".
- If a duplicate exists, identify and the duplicate work item ID

6. Add Skills (pro tip, you can search and select multiple skills and click 'add' once)

  • Search with JQL

  • Update labels

  • Update work item

  • Update status

  • Find similar work items

  • Comment on work item

  • Get work item


7. Configure Knowledge, Select 'custom knowledge', click 'add knowledge'

  • Select Confluence, select spaces and select the space where you saved the Engineering Triage & Backlog Policy page.

  • Select Jira and and choose your team space


  1. Conversation starters:

    1. Analyze my backlog for stale issues

    2. Find duplicate work items in my backlog

    3. Show me items missing key information


  1. Web search: toggle off

  2. Deep research: leave off

  3. Select 'test'

  4. Select Publish' Remember, the update action in rovo is to update the comments


Step 4: Hook Jira Automation to the Agent

Now we set up the trigger. Instead of running this manually, let Jira kick off the cleanup sequence every Monday morning.

  1. In Jira, go to Project settings > Automation > Create flow > create from scratch

  2. Add a trigger: Select Scheduled (e.g., Monday at 6:00 AM). Ensure the option to Run a JQL search and execute actions for each issue in the query is enabled.

  3. Select run a JQL search, paste this JQL query, remember to replace with your project name

    project = "YOURPROJ" AND status in ("Backlog", "To Do") AND updated < -90d AND resolution is EMPTY


  4. Action: Select Rovo Agent and select your backlog hygiene specialist agent

  5. In the message field, write:

    Issue: {{issue.key}} — {{issue.summary}} Current Description: {{issue.description}}


    Task: 1. Compare this issue against the "Definition of Ready" in the Confluence policy page in your knowledge base. 2. Identify any missing elements: user story format, acceptance criteria, epic link. 3. Add an internal comment starting with "🤖 Rovo Triage Evaluation" listing what is present and what is missing. 4. If the issue has not been updated in 90+ days, add the label "stale-candidate". 5. Apply the label "rovo-reviewed" to all issues you process. Do not transition or close any issues. Comment only.


  6. Click Turn on flow.


Step 5: Add Quick Filter to Team Board

Never let an AI automatically delete or transition your tickets without human sign-off. That is how important work gets lost. Instead, create a review dashboard.

  1. As a board admin, go to your Jira Team backlog board and add a new Quick Filter.

  2. Title it: Requires Cleanup Action

  3. Enter this JQL criteria:

    labels in ("stale-candidate") OR (labels = "rovo-reviewed" AND description !~ "Acceptance Criteria")

  4. During your 15-minute weekly grooming synch, click that filter.

  5. Review Rovo's comments in each of the work items, confirm they are correct, and take action.


Proving It Works

Once your automation executes next Monday:

  • Look at the Jira History: You must see automated comment updates styled with markdown, outlining completeness summary.

  • Label Check: Stale issues should instantly show the label stale-candidate.

  • Audit Trail: Check Project Settings > Automation > Audit Log to verify that Jira successfully reached out to the Backlog Hygiene Specialist and received API responses without timing out.


Thats it, I hope this was helpful to you

Comments

bottom of page