Events are the Wrong Abstraction for Your AI Agents

Explore why event-driven architecture may not be the optimal choice for building AI agents in this 15-minute conference talk. Learn about the inherent challenges of using events as communication abstractions in distributed AI systems, including fragmented logic, increased latency, and decreased observability. Discover how durable execution provides a superior alternative that delivers the benefits of event-driven architecture while eliminating its complexities. Examine the specific pitfalls of event-driven patterns when applied to AI agent development and understand why durable execution represents the correct abstraction for building robust, scalable AI agents that need to communicate with tools, data repositories, and other agents across networks. Gain insights from a senior developer advocate who specializes in distributed systems and has extensive experience in building developer-focused educational content and community engagement.