Andrew Guenther
"The more things change, the more they stay the same."
As models become increasingly capable of debugging complex systems, building effective operational agents starts to look a lot like building effective engineering teams. How do we ensure people and agents have the information they need? How do we help them collaborate? How do we build trust as agents evolve from answering questions to making proactive suggestions and, eventually, taking autonomous action?
Drawing on our experience building operational agents at OpenAI, we’ll explore how to make agents effective partners for responders.

Andrew Guenther is an engineer in OpenAI's infrastructure organization where he leads work on reliability and incident response systems. He runs OpenAI's high-severity incident commander rotation and builds agentic tooling to help engineers better build and operate their services. Andrew has spent more than a decade working on DevOps, cloud infrastructure, and production reliability.

