We deployed AI agents across our company, including autonomous support in Jira. AMA.
About us. We build field staff automation software for SMBs. Our core stack is Jira, Confluence, GitLab, Graylog, and Telegram for customer communication.
The initial trigger was to cover support gaps during nights and weekends. We also wanted an internal knowledge assistant the whole team could use to get answers that usually require a developer. Finally, we wanted to automate writing technical and user documentation.
We didn’t want to end up with a separate AI solution in every system, each with its own setup, overhead, and cost. So we settled on an independent AI workspace, and implemented all our agents there:
- external support agent that works in Jira like a regular rep,
- internal assistant available to every team member,
- docs writer,
- specialized Graylog miner every other agent can call as needed.
Results so far:
- 70%+ auto-resolve on repetitive L1/L2, response time in seconds vs hours. A lot of room to grow, as we review AI responses weekly and adjust instructions/knowledge.
- Internal assistant use is rivaling external one. Team found it very helpful, especially because it pulls data from so many sources.
- Tech docs writing deployment is ongoing.
- We're now experimenting with an agent that reviews the work of other agents and suggest fixes.
A bit about the external agent flow. The agent connects to Jira as a regular user and handles tickets directly:
- Triggers on assignment, reads ticket + comments + attachments,
- Writes responses, updates status, escalates when confidence is low,
- Pulls knowledge from Confluence, internal docs, APIs, observability logs,
- Hard-scoped to current ticket, respects Jira permissions, shows up in audit logs,
Happy to answer any questions.