Abstract
What if AI agents could support your entire development workflow, not just code generation? In this session, Benedikt will walk through a practical approach to deploying specialized agents across the full software development lifecycle: from refinement and planning, through coding and review, all the way to architecture validation and retrospectives.
We'll explore how to maintain human-in-the-loop control while letting agents handle the heavy lifting, and discuss the cultural shift teams need to make this work. Expect whiteboard sketches, real examples, and plenty of room for questions.
Three takeaways
- Every step of the lifecycle runs the agent's own loop of research, plan, implement and reflect, so a team's whole process can become a pipeline of agents.
- Let the agent loop over its own first implementation, and drop the code review once its results hold up.
- When you correct the agent, have it write down why, and let an agent run the retrospective over the sessions to improve the pipeline.