AI Isn't Replacing SREs. It's Deskilling Them

Your ability to troubleshoot novel system failures degrades as AI handles routine incident response.

The article posits that Artificial Intelligence (AI) is not eliminating Site Reliability Engineering (SRE) roles, but rather deskilling the engineers performing them. Drawing parallels to industrial automation's impact on blue-collar workers, the author argues that as AI handles the majority of routine incident response and system monitoring tasks, human SREs are left with fewer opportunities to practice and develop expertise in handling rare, complex, or novel outages. This phenomenon, identified in research by Lisanne Bainbridge in 1983, suggests that increased automation paradoxically makes human operators less capable during critical failure events due to a lack of practice. Current AI in SRE automates alert reduction, runbook execution, log analysis, and basic root-cause suggestions, with future advancements targeting more complex diagnostics and predictive incident detection. The concern is that SREs may not realize their skills are atrophying until a critical, unforeseen incident occurs, a risk amplified for junior engineers who may never develop these foundational skills, termed 'never-skilling'.

The core irony of automation, as observed in industrial settings, is that it renders human operators more critical during system failures while simultaneously reducing their practice and thus their proficiency. AI is now replicating this dynamic within knowledge work, including SRE, by automating common tasks and leaving humans to handle the increasingly rare, but high-stakes, exceptions for which they may be undertrained. This creates a growing gap between perceived productivity gains and actual latent capability, potentially leading to significant issues during unforeseen incidents.

The risk extends beyond deskilling, impacting junior engineers who may experience 'never-skilling' by not developing foundational expertise due to AI's preemption of hands-on incident response. Companies must actively design for human engagement, potentially through 'deliberate inefficiency'—scheduling human-led practice on AI-handled incidents—to maintain and develop critical skills. Observability platform providers like SigNoz are exploring ways to integrate AI while ensuring human control, leveraging their direct data access for more context-aware AI assistance, and aiming to build trust through phased autonomy.