AI Ticket Automation
The project began as a Go-based proof of concept designed to validate whether an AI agent could take a Jira ticket from assignment to implementation. The service retrieved eligible tickets, invoked Hermes to execute the development task, and automated the creation of branches and pull requests in Bitbucket. This POC demonstrated that the workflow was viable, but it also revealed an opportunity to simplify the architecture. Rather than maintaining a separate orchestration service, the workflow could be moved directly into Hermes using scheduled agent jobs.
Technologies
Problem
Low-risk, repetitive engineering tasks still require developer time to triage, implement, review, and maintain through the pull request lifecycle. The initial goal was to explore whether AI-driven automation could reliably handle these tasks while fitting into the team’s existing Jira and Bitbucket workflows.
Solution
Developed an AI-driven engineering workflow using specialized Hermes agents and scheduled jobs to automate low-risk development tasks throughout the pull request lifecycle. The system integrates with Jira and Bitbucket to discover eligible work, implement changes, create and maintain pull requests, address issues, and participate in code review—while keeping the workflow within the tools already used by the engineering team.
Contributions
- 01 -
Developed the original Go-based proof of concept to validate autonomous Jira-to-pull-request workflows.
- 02 -
Evaluated the POC and helped evolve the architecture from a standalone orchestration service to native Hermes Agent workflows.
- 03 -
Designed and configured specialized scheduled jobs for ticket discovery, implementation, PR creation, synchronization, bug fixing, and review.
- 04 -
Integrated Jira and Bitbucket into the agent workflow to connect task management with the existing development process.
- 05 -
Built iterative feedback workflows that allow agents to respond to PR comments, address bugs, and revise previously generated changes.
- 06 -
Explored observability and performance metrics for agent executions, revisions, failures, model performance, and pull-request feedback.