Back Back
AI agent for bug prioritization
Prioritizes every new Jira bug, alerts the PM when it’s critical and is checked with evals.
- What it is
- Every time a bug is created in Jira, an AI agent analyzes it, classifies it and assigns a priority from 1 (very critical) to 5 (trivial). If it’s critical, it alerts the PM right away.
- Problem it solves
- Prioritizing bugs by hand is slow and the criteria change depending on who reviews them, so a critical bug can get buried in the backlog. The agent prioritizes them consistently and alerts as soon as something critical comes in, so the team can react fast.
- Who it’s for
- PMs and development teams managing bugs in Jira.
- How I built it
- An AI agent connected to Jira. Its accuracy is measured with evals against a golden dataset of bugs with their expected priority.
- Built with
- Jira, AI agent, evals, HTML email
- What I delivered
- Automatic priority from 1 to 5 for every new bug.
- An instant HTML email for critical bugs, with ID, title, description, urgency, type, project, reporter, date and a direct link to Jira.
- Evals against a golden dataset to measure accuracy.
Walk through the workflow
Step 1 of 5
Someone files a new bug in Jira.
Step 2 of 5
Reads the bug and classifies it.
Step 3 of 5
From 1 (very critical) to 5 (trivial).
Step 4 of 5
If so, it moves on to the alert. If not, the bug keeps its assigned priority.
Step 5 of 5
Sends an HTML email right away with the key bug details and a direct link to Jira.
How it’s measured
The agent’s accuracy is measured with evals against a golden dataset of bugs with their expected priority.
On the 8-bug golden dataset, the agent assigned the expected priority in 7 of 8 (87.5%).
Evidence

Workflow in n8n: production and evals 
Golden dataset: expected priority vs. the one the agent assigned 
Alert email to the PM about a critical bug 
Jira backlog 
Eval runs in n8n