A QA team faced a high-risk task involving complex data flows across multiple microservices. The project required analyzing over 8,000 documents in Confluence to ensure complete test coverage within a strict deadline.
The team implemented a multi-agent AI process to handle the analysis. They broke the task into specialized sub-tasks, assigning each to an AI agent with clear instructions, defined input-output formats, and specific context constraints.
The process involved two main stages: identifying requirements for Kafka topics and SQL databases, and mapping these against the actual codebase. This generated a comprehensive report linking documentation to code, which was then verified by developers and analysts.
The team transformed this methodology into a reusable environment for Cursor AI agents. This tool now helps the team identify implementation defects, clarify functional requirements, and explain service behaviors by linking production code with documentation.
The environment allows agents to answer questions about expected versus actual system behavior, analyze new Git branches for discrepancies, and update the knowledge base while maintaining references to original sources.
