Note: An in-team initiative; process and tool names are kept generic.
Summary
We built an AI agent that acts like a business analyst inside our team. It is triggered automatically when a task is assigned to it or when a /analiz comment is added; it scans the related epic, Confluence pages and connected code repositories to make the task scope clearer, more structured and more actionable.
Problem
Task scope was often unclear, and developers spent significant time on analysis before starting. Inconsistent, incomplete scope slowed down and made unpredictable our AI-powered (“vibe coding”) development flow.
Constraints
- Scope had to be clear, structured and actionable.
- It had to integrate with existing tools (task management, Confluence, code repositories).
- The process had to be transparent and traceable — not a black box.
Decision
We decided to automate analysis so it was no longer a repetitive, person-dependent bottleneck: a main agent acting as a business analyst, supported by sub-agents and dedicated skill sets triggered by a task number, and a plan-mode driven development flow.
Implementation
- The agent triggers automatically on task assignment or a
/analizcomment. - It scans the related epic, Confluence pages and connected code repositories.
- It researches the required changes and enriches task scope (clear, structured, actionable).
- Sub-agents and skills triggered by a task number generate a plan-mode driven development flow.
Outcome
Developers start with a well-defined scope, spend less time on analysis, and move into implementation much faster. The path from analysis to implementation became more transparent, traceable and efficient, and AI-assisted development accelerated significantly.
Lessons Learned
Investing in in-team tooling and agents pays back by compounding delivery speed. Combined with process visibility and repeatability, automation raises both speed and quality together.