Webinar: Why Agentic AI Projects Fail

A webinar by Thiago Grabe examines why most agentic AI projects break down once they leave the demo stage, and what it takes to build production-ready systems that hold up under real-world conditions.

The gap between a working demo and a deployed agentic system is wider than it looks. In this session, Thiago Grabe walks through real-world failure cases and distills practical lessons for designing agentic AI systems that survive outside controlled environments.

The webinar covers why successful demos so often fail to translate to production, the hidden challenges of running agentic systems at scale — including cost, observability, and guardrails — and the architectural patterns that work, along with their tradeoffs. The discussion treats reliability, control, and overall system design as core concerns rather than afterthoughts.

For researchers, engineers, and PhD candidates working at the intersection of LLMs, agent frameworks, and applied systems, the session offers a grounded perspective on what is actually difficult about deploying agentic AI, along with a useful vocabulary for thinking about those difficulties.