On-Call Rotation Design for AI Agent Platforms
Human judgment of agent decisions must become an explicit SLO metric in on-call rotations.
Yuki Tanaka-Osei
Section
5 stories in Incident Management.
Human judgment of agent decisions must become an explicit SLO metric in on-call rotations.
Agent failures demand a new postmortem structure that tracks both infrastructure and reasoning.
Autonomous systems need new severity rules built around damage that can't be undone.
Detect, contain, and roll back all four artifact layers together.
Fast detection and recovery paths for OpenClaw's most destructive production failure modes.