Event Storming facilitation is becoming a new hard skill, but learning it is painfully slow. You only get better by running workshops, and most engineers don’t get to run many - often in the same domains, with the same goals and with the same people over and over again. That’s definitely not enough to develop real skills.

So we asked a question:
What if Event Storming participants weren’t people at all?

In this talk, we’ll walk through our case study of building an agent-driven Event Storming learning tool using LangGraph. The core challenge turned out to be multi-agent collaboration: how agents communicate, coordinate and work together toward a shared goal. We’ll show which communication structures worked for us, where collaboration patterns helped and what failed along the way.

We’ll also briefly touch on how we observe and debug such systems in practice, using real examples from a running multi-agent setup.

After the talk you’ll know:
- How to design agents that behave like real workshop participants
- How to choose multi-agent communication structures for a concrete problem
- Context engineering techniques that matter in long-running collaborative sessions
- Battle-tested agentic design patterns
- How to observe and debug agent-driven systems

See you!