AI agents are already making decisions on behalf of product teams. Some of them are doing it badly, and most product leaders don't know it yet.

Csaba Tamas delivers a practical field guide for product leaders building or scaling agentic systems in production. He starts where most projects go wrong: ROI selection. The highest-return agent projects are unglamorous, repetitive, and operationally well-defined. Knowing the difference before committing is the first competitive advantage.

He tackles what silently kills most agent projects: what does 'working' actually mean? Agents fail softly in ways traditional software never did. Csaba establishes task success rate as the foundational metric, and explains why a ~73% success rate in a customer-facing workflow isn't a beta problem. It's a liability. He covers the Agent Mesh, observability, and why governance must come before scale. Plus the uncomfortable truth about RAG: most failures are curation problems, not engineering ones.

Attendees leave with a framework for Monday morning: how to select the right agentic investment, instrument it properly, govern it responsibly, and know, with evidence, whether it's actually working.