29 July 2026 09:00 - 09:30
Why agent systems fail in production
As agents interact with tools, APIs, data sources, and other agents, small issues quickly compound into unreliable behaviour, unpredictable outcomes, and systems that are difficult to monitor, debug, and trust.
This keynote explores the most common reasons agentic systems break down after deployment, from orchestration failures and poor observability to evaluation gaps, tool reliability, and escalating complexity.
We'll examine what separates successful production deployments from promising prototypes, and the architectural and operational lessons emerging from teams running agents at scale.
Attendees will leave with a clearer understanding of where agent systems fail, why those failures occur, and how to build more reliable, observable, and production-ready agentic applications.