Google
S11
HardPremiumDesign an AI On-Call / Incident-Response Agent
Design an AI agent that helps on-call engineers respond to production incidents — investigating alerts, correlating logs and metrics, identifying root cause, and suggesting fixes.
AgentsTool useApproval gatesObservabilityReliability
Key Requirements
- Start investigating within seconds and give a grounded summary within about 2 minutes (assumed target)
- Correlate logs, metrics, traces, deploys and config changes; cite evidence for every claim
- Never make things worse: gate writes by reversibility and blast radius, verify and roll back
- Explain its reasoning with evidence the engineer can check
- Handle alert storms and cascades, and keep working when its own dependencies fail
Interviewer Follow-ups
- Q1How do you prevent the agent from executing a dangerous fix?
- Q2How does it handle cascading failures across services?
- Q3What if the agent's own infrastructure is down during the incident?
Loading...