Agentic documentation feedback pipeline
How I reduced resolution time on docs feedback by 97% with an agentic pipeline that triages and actions customer feedback.
Tools + Tech
Cursor
Jira
Markdown
Atlassian Rovo MCPTens of thousands of developers visit Atlassian's docs every month.
Many leave feedback — flagging broken links, outdated endpoints, missing examples. For years, those tickets piled up untouched.
I assembled a team of 4 content designers to run weekly triage sessions, reviewing and actioning the backlog. It worked, but it wasn't scalable — across the four of us, we were spending 8–10 hours a week of combined effort just to keep up.
I built an agentic pipeline in Cursor to automate how we action customer feedback.
I codified my approach to triaging and updating docs as a technical writer, providing instructions to Cursor as a custom agent and connecting to Jira via the Atlassian Rovo MCP. Here's how it works:
What used to take ~8-10 working hours each week now takes around 30 minutes on a Monday morning.
The pipeline has:
- Reduced feedback ticket resolution time by 97%
- Resolved over 300 tickets in 6 months
- Lifted page sentiment by 20% (3.0 → 3.6 avg rating)
Improvements are iterative. Next things I'm hoping to tackle are: auto-routing tickets to the right engineering team when SME input is required, richer feedback capture with issue categories and contact details, and self-healing pipelines that proactively detect issues like broken links before users even notice.