When Incidents Fix Themselves: AI SRE in action
summary
Claim: AI SRE agents diagnose and remediate outages autonomously by combining observability, deployment context, and code access. Demo shows a root cause analysis in 2 minutes vs 10, a Kubernetes memory patch applied, a rollback option, an auto-created pull request and Kubernetes patch, and an updated status page. Architecture uses owned MCP servers, an LLM orchestration layer, schema-driven JSON output, ~20 tools in the loop, and RCA runs that can consume 200k–300k tokens, yielding consistent 10x MTTR improvements in many cases.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 4:00Why agents are SRE’s superpower
- 9:23Live demo: AI root cause analysis
- 19:22Key learnings from RCA agent
- 29:04Agent architecture and MCP server
- 34:53Future roadmap for decentralized agents
- 36:39Q&A
related talks
30mAI Agents for Incident Management
36mSRE'ing the SRE AI Agents
16mHow AI is redefining SRE and Customer Experience
44mFrom 4 Hours to 8 Minutes with AI Agents That Transform SRE...
15mIntelligent Incident Management in DevOps
35mAgentic Incident Response
05mAI-Driven Incident Resolution – Hype or Reality
43mCONTEXT ENGINEERING FOR SELF-HEALING AI SRE