AI Eases Incident Response While Eroding the Skills SREs Need Most
AI-assisted incident response risks deskilling SREs, argues SigNoz, citing automation research and medical and aviation examples, though the evidence is analogical and remedies untested.
SigNoz's Elizabeth argues that AI is not replacing site reliability engineers but deskilling them, drawing on Lisanne Bainbridge's 1983 Ironies of Automation to warn that automating routine incident work leaves humans less practiced for rare, high-stakes failures.
Current AI-assisted SRE tasks include alert grouping, runbook execution, log anomaly flagging, root-cause suggestions, and auto-remediation, with PagerDuty's tiered model reserving novel cascading failures for humans. Evidence includes a study finding endoscopists' unassisted polyp detection fell from 28% to 22% after AI use, and aviation autopilot degradation that prompted FAA manual-flying mandates.
Junior SREs may never develop foundational intuition, and proposed remedies include deliberate inefficiency and human-in-the-loop engagement. SigNoz says its AI assistant keeps humans in control, with more autonomy only as trust is earned.
Key takeaways
- Bainbridge's 1983 paper found automation leaves operators with passive monitoring and rare crises they are increasingly unprepared for, a pattern applied to AI in SRE.
- Current AI-assisted SRE work includes alert noise reduction, runbook execution, log pattern detection, anomaly flagging, root-cause suggestions, and auto-remediation; PagerDuty's model assigns Tier 1 known issues to full automation, Tier 2 to AI recommendations with human validation, and Tier 3 novel failures to humans.
- A study found endoscopists using AI polyp detection saw unassisted detection drop from 28% to 22%, and aviation research shows autopilot reliance degrades situational awareness and manual flying, leading the FAA to mandate more manual flying time.
- Junior SREs may never develop foundational intuition if AI handles most incident response, creating a cycle where skill degradation leads to more AI reliance.
- Proposed remedies include designating some AI-capable incidents as human-practice opportunities and building human-in-the-loop systems where humans actively engage rather than merely approve AI recommendations; SigNoz plans an AI assistant with humans in control and later autonomy within guardrails.
What people think
Analysis details
24 comments analyzed. Analyzed .
The sample selects active discussion branches and includes available parent comments. It may omit parts of the full thread. The analysis included 24 of 37 usable stored comments because of the input limit. Selected themes do not measure community opinion or how common a view is.
SRE role blurs into platform engineering
The label covers everything from Google-style engineering teams that co-design systems and can throttle releases to Kubernetes-and-Terraform platform roles. If SRE is just rebranded ops, the deskilling question may concern a job title already in flux rather than a stable profession. ZIRP-era DevOps is criticized for creating separate groups that by definition were not DevOps, and some argue tool-specific skills age poorly compared with foundational ones.
AI reliance erodes hands-on skill
Self-reported experiences describe coding ability and motivation atrophying as AI handles more work, with interview readiness suffering. A counterargument says new skills in prompting, routing output, and spotting hallucinations replace the old ones. A reply challenges that: spotting hallucinations requires the domain knowledge that hands-on work builds, so if AI does all the work, the knowledge never forms. This makes the deskilling risk a training-pipeline problem, not just a productivity tradeoff.
Aviation analogy faces incentive and scope doubts
Aviation is cited as a model for mandated manual practice, but the analogy draws skepticism. Aviation's safety rules may be enforced by legal hedging and rent-seeking that software lacks, and a pilot argues newer jets push single-pilot operations, undermining the manual-flying premise. Others note zero fatal mistakes is impossible and not every system needs aviation-grade reliability. This questions whether deliberate inefficiency can transfer without similar regulatory pressure.
Zero-mistake reliability carries a cost
Aviation-grade reliability is described as achievable only at a cost, and not all projects need the same properties; some systems can tolerate scheduled maintenance. At the same time, production data loss and security breaches are areas where teams do try to avoid mistakes. This qualifies the deliberate-inefficiency proposal: practice time and redundancy consume resources, so the tradeoff depends on the system's criticality.