AI handles incidents, engineers lose touch with their systems
Points and comments are a snapshot, not live.
AI incident response tools risk deskilling engineers for complex failures.
Sylvain Kalache, an SRE at LinkedIn in 2012, now works at Rootly on incident management. He warns that AI SREs, which automate routine incident response, prevent engineers from developing intuition about system failures. Citing Lisanne Bainbridge's 1983 paper "The Ironies of Automation," he argues that automation leaves humans responsible for novel crises while reducing practice. He predicts average MTTR will drop but resolution times for complex incidents will rise. Kalache advocates realistic incident simulations, like Rootly's partnership with Uptime Labs, to maintain skills. He compares the risk to aviation, where pilots regularly rehearse rare emergencies in simulators, contrasting with engineering teams that lack such drills.
What commenters are saying
Commenters largely agree with the article's premise but split on the severity. A top comment argues that if capability growth continues, AI-unsolvable incidents will stump humans regardless of practice, citing how engineers rarely hand-write code anymore. Others counter that human operators can still use AI as a tool to triage novel failures, though this advantage may fade for post-AI hires. A subthread draws an analogy to third-party libraries: losing intuition for AI-generated code is like using black-box dependencies, risking technical debt. One commenter reports seeing team members regress in troubleshooting skills, relying on AI outputs without understanding root causes.