Listen to Production the Way It Deserves
summary
Traditional alerts are guesses based on thresholds and they cause noise and toil, so the solution is agentic AI that correlates telemetry, logs, code and Slack via an MCP but fails without production context. Running continuous cartographer agents builds a live production map that boosts root cause accuracy from 35% to 70% in experiments and cuts investigation time from over 15 minutes to under 4 minutes by producing correct log queries and service semantics up front.
AI-written summary · about
chapters (11)
- 0:00Welcome
- 1:07Production deserves attention
- 2:32Limitations of traditional alerts
- 4:38Hearing vs listening to production
- 6:24AI SRE agents challenges
- 9:58Context problem for AI agents
- 13:10Introducing cartographer agents
- 17:21How cartographer agents work
- 19:52Future with AI-powered maps
- 21:14From firefighter to navigator
- 22:53Q&A
related talks
21mAgentic Monitoring in Production
27mAI-Powered Reliability: Stop Fighting Fires, Prevent Them
15mIntelligent Incident Management in DevOps
18mProduction software keeps breaking and it will only get worse
25mSymptom-based Alerting for Machine Learning - What I Learned...
27mAI-Powered Root Cause Analysis at Scale: From Theory To Production...
08mProduction Evals For Agentic AI Systems
35mAgentic Incident Response