Modern cloud-native infrastructure generates more alerts than human engineers can ever realistically handle. We are stuck in a loop of getting paged, checking Grafana, and executing runbooks. It’s time to move past static alerting and build systems that can actually fix themselves.
This session bridges the gap between traditional infrastructure-as-code and Generative AI for Day-2 operations. I will demonstrate how to build an "Agentic SRE" workflow that doesn't just page you—it investigates. We will walk through a practical, production-ready event-driven architecture that catches Prometheus webhooks, passes the payload to an LLM to query open-source telemetry for context, and finally proposes safe infrastructure-as-code fixes via Terraform and kubectl. This isn't sci-fi; it's just smart orchestration. You will leave with a clear roadmap for implementing AI-driven incident response in your own high-scale Kubernetes environments.