AI may cut routine incident times while leaving engineers less prepared for complex failures
An article warns that automating software incidents could reduce average recovery times while eroding the hands-on experience engineers need when systems fail in unfamiliar ways.
AI-assisted incident-response tools can inspect alerts, form hypotheses, query telemetry, correlate recent deployments and implement fixes, according to an article published on 5 September. The author argues that these capabilities could lower mean time to resolution (MTTR) for routine software incidents.
The concern is that routine incidents also give engineers practical experience of how systems behave and fail. As automation handles more of that work, responders may become less familiar with the systems they oversee and struggle when an ambiguous, high-severity incident falls outside an automated tool’s capabilities.
The author predicts a widening gap between average and complex-incident performance: routine recovery times may fall, while resolution times for unusual failures could rise. The article presents this as a forecast, not an independently measured result, and does not quantify any loss of system familiarity.
The argument draws on human-factors researcher Lisanne Bainbridge’s 1983 paper, “The Ironies of Automation”, which held that automation can reduce opportunities for routine practice while leaving operators responsible for abnormal situations. Aviation is offered as an analogy, with pilots using simulators to rehearse rare emergencies that automation cannot manage.
Rootly, where the author works, and Uptime Labs are cited as examples of incident-response training using simulated outages, observability tools and LLM-powered stakeholders. Engineers take the incident commander’s role during a simulated e-commerce outage, practising investigation, communication and coordination. The example is self-reported and is not independently validated in the article.
The author recommends regular hands-on work, tabletop exercises, simulations, chaos engineering and emergency rehearsals. AI explanations of its decisions may help responders understand an incident, but the article argues that observation cannot replace practical experience.
