An opinion piece traces the evolution of production support from developer-led troubleshooting to specialized L1/L2/L3 teams to today's AI-powered SRE approach. The author argues observability alone is insufficient given the volume of telemetry in modern distributed systems, and proposes AI as a correlation and reasoning layer that narrows down probable problem areas using specialized 'skills' (Kubernetes, database, API, platform) rather than fixing issues outright. It emphasizes combining live telemetry with historical incident knowledge, measuring value beyond binary resolution (e.g., time-to-direction), and evolving toward 'performance intelligence' focused on user experience rather than just infrastructure metrics. The author envisions a progressive path from observation to correlation to recommendation to validation to eventual automated remediation, with engineers remaining central to judgment.
