LLMs Broke the SRE Runbook. Now What?

Reliability engineers are the quiet force that keeps modern software running. After decades of refining practices on deterministic systems, many teams are chasing yet another “nine” beyond 99% uptime. But the AI era, especially LLM-backed features, changes the game. Outputs are non-deterministic, data pipelines shift underfoot, and key components behave like black boxes. As a result, many of the tools and rituals SREs have mastered for decades no longer map cleanly to production AI.

At SREcon EMEA 2025, I co-organized the MLOps discussion…