On-Call the Easy Way With Agents
summary
Intuit reports a deployed Agent called Mesh Debug Deputy that ties PagerDuty alerts to Splunk and Wavefront, RAGs over internal docs, and tool calls like kubectl/istioctl to automate on‑call steps. It runs across ~350+ Kubernetes clusters with ~2,000 rollouts, ingests ~16 billion daily transactions and 292k peak DPS, and claims reduced MTTR by letting the agent fetch logs/metrics, choose runbook steps, and either suggest or dry‑run fixes.
AI-written summary · about
chapters (13)
- 0:00On-call with AI agents
- 1:24Intuit and its scale
- 2:50Debugging at scale
- 4:15AI agents for on-call
- 6:46Agent architecture
- 8:12Retrieval augmented generation in action
- 10:18RAG knowledge base
- 12:05Tool selection role
- 14:32Tool registry
- 14:54Runbooks to tools
- 17:02Live debugging demo
- 18:27Agent benefits and goals
- 20:59Q&A
related talks
28mThe Agent That Turned On-Call Chaos into a Single Slack Thread
28mOn-Call Experience at Uber
30mLive Panel with
13mAppDynamics - Path to Kubernetes Observability
35mThriving With Kubernetes On-Call: Best Practices & Lessons Learned
46mObservability in the real world: Improve systems w/iterative approach
19mAI Agents for Incidents: The Good, the Bad, and the Ugly
44mFrom 4 Hours to 8 Minutes with AI Agents That Transform SRE...