The magic of AI means you can now build a useful agent in an afternoon. But the path from desktop demo to production platform is treacherous, littered with the careers of CIOs who underestimate the caprices of our new LLM overlords.

In this talk, I’ll discuss the challenges we as an industry face as we boldly attempt to incorporate autonomous AI agents into the innermost sanctums of our private infrastructure. I’ll show that AI agents don’t so much change the game we’re playing as raise the stakes for success and failure.

With that insight in hand, I’ll show how the battle-tested streaming approach to building secure, scalable, and maintainable event-driven applications holds even more relevance in the brave new world of agentic AI. As we all strive to keep pace with the fast-changing landscape of AI frameworks and development tools, streaming provides a consistent, foundational approach to agentic infrastructure that applies broadly and uniformly across them all.