No More Tough Nights: How AI Transformed Our On-Call Duty
summary
Iguane Solutions presents nAIghts Watch, a multi-agent AI system that automatically handles level-1 on-call incidents (disk full, unresponsive service, high CPU) by detecting alerts, analyzing logs and metrics, running pre-approved SRE commands (restart, log rotation, etc.), verifying recovery, and generating a full incident report. Deployment of the system reduced on-call wake-ups by 80%, sped up resolution from about 60 minutes to ~3 minutes, and freed SRE time for innovation and further training of the agents.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 2:13On-call incidents daily
- 6:53Nightwatch AI project
- 12:30Single agent limits
- 14:12Three agents judge solution
- 16:42Demo: Nightwatch CPU incident
- 18:47Securing server access
- 21:15Log token optimization
- 24:29Time and cost results
- 27:38Team and customer impact
- 29:02Improvements and future work
- 34:19Nightwatch lessons
- 36:24Summary
- 38:51Q&A
related talks
24mAI and Developers: Revolutionizing Problem Solving
11mAI-Native Network Operation
39mAI as a Catalyst, Not a Replacement
19mFrom Alerts to Action: AI-Powered Incident Response Systems
30mAI and Incident Management: How to Reduce Manual Work and Speed up the Incident...
42mAI for security - friend or foe?
15mIntelligent Incident Management in DevOps
15mAI Assistance for On-Call Engineer