The Rise of Agentic SRE: Humans, Agents, and Reliability

As organisations deploy increasingly complex and distributed architectures on Google Cloud spanning GKE, Cloud Run, Compute Engine, Cloud SQL, Spanner, Pub/Sub, and multi-region networking, traditional incident response based on reactive practices falls behind. The need arises for smarter automation that enables faster anomaly detection, richer context analysis, and timely interventions within well-defined operational constraints. In this talk, I will discuss the principles of building a robust and dependable platform for agentic SRE on Google Cloud, the key architectural components, and the observability layer that provides the foundational data. I will also examine the safeguards required to enable autonomy while keeping operations secure, governed, and auditable. Finally, through practical use cases alert triage, incident copiloting, root-cause analysis, and bounded automated remediation, we will explore how AI can be integrated into operations to improve reliability.