An observability vendor warns AI incident response will leave SREs unable to fix novel outages

Bainbridge's 1983 paradox, applied to the pager

The essay on the SigNoz newsletter argues that AI in site reliability engineering may quietly erode the skills it was meant to supplement. The worry is the rare, novel incident, which the author calls "the 5% that actually matters". The foundation is Lisanne Bainbridge's 1983 "Ironies of Automation": the more a process is automated, the less practice operators get, though they are most needed when the automation fails.

PagerDuty's tiered model makes the problem concrete. Known issues are automated at Tier 1, AI recommends fixes for partly understood problems at Tier 2 with humans validating, and novel failures stay human-led at Tier 3. An engineer who only ever sees Tier 3 never builds the intuition Tier 3 demands. The essay separates deskilling (losing skills you had) from "never-skilling" (juniors who never acquire them), and adds that AI makes work feel easier, so people overestimate their competence and lean on it more.

The evidence is borrowed: a Lancet study in which endoscopists' unassisted polyp detection fell from 28% to 22% after a period of AI use, and aviation research tying heavy autopilot use to weaker manual flying, which is why the FAA now mandates manual flying time. The remedy is "deliberate inefficiency": mark some incidents, even ones the AI could resolve, as human practice. SigNoz sells observability tooling and says its assistant keeps humans in control, so the framing is not neutral.

"Root cause in a couple of minutes" versus "skill atrophy is real"

The strongest rebuttal came from people using the tools on production. An engineer who gave an LLM a token to call the Jenkins API described the before and after:

"I used to scour through server logs, build logs, test reports etc to figure out why tests were failing in CI. That was exhausting work and could take days of investigation. Now I get the exact root cause in a couple of minutes."

A startup engineer who covers the whole stack agreed on the speed, then conceded the essay's point about themselves:

"skill atrophy is real. Thinking about getting back on the interview circuit is p scary because of it."

The calculator analogy, and why aviation may not transfer

One commenter, cadamsdotcom, argued that nobody mourns long division. debo_ answered that the stakes differ:

"I don't think your inability to do long division will cause as much damage as an inability to bring the site back up in an emergency situation."

The FAA comparison drew its own pushback. nrmitchi pointed out that the practice is mandated and uniform across providers, while "Modern companies generally do not have the same incentive to keep their SRE (or engineers in general) upskilled." The founder of a startup that sold training through simulated outages was blunter about demand:

"without the FAA mandating something like this for SRE's, no one will do it."

His closing line, "Maybe we were too early", is the most telling data point in the thread. Drills like this already exist inside some teams: one commenter described a standing "Wheel of Misfortune", where a facilitator breaks a non-prod environment, spins a wheel, picks a victim, and the team solves it manually. The open question is which employers will pay for that time.