The article argues AI isn't replacing SREs but deskilling them by automating routine tasks. Elizabeth from Sig Noz discusses how AI now handles alert noise reduction, runbook execution, log pattern detection, and basic root-cause suggestions. This mirrors the 1983 'Ironies of Automation' paper by cognitive psychologist Lisanne Bainbridge, which found that increased automation makes human operators more critical during failures yet less practiced. Today's SREs face similar risks: AI manages well-understood failures like restarting crashed pods, leaving humans with rare, complex problems requiring deep expertise that atrophies without practice. The author draws parallels between industrial automation reshaping blue-collar work and AI transforming knowledge work. While AI agents improve, SREs lose muscle memory for critical interventions. The piece questions whether anyone will notice the capability decline before the next major outage hits.
Pager Duty frames incident response in tiers: Tier 1 incidents with known fixes get fully automated, Tier 2 receives AI recommendations with human validation, and Tier 3 novel cascading failures remain human-led. But the core concern remains: as AI handles 95% of incident response, do SREs get worse at the 5% that matters most?
The foundational framework comes from Bainbridge's research showing automation creates paradoxical outcomes - humans become more essential during automation failures precisely when they have the least practice. This pattern now repeats with AI agents in software systems, where stakes continue growing.
Source: Hacker News · Summarized by HeadlinesBriefing