Site Reliability Engineering is a discipline that applies software engineering practices to operations work. SRE emerged at Google in the early 2000s and was popularized through the 2016 book "Site Reliability Engineering" (O'Reilly). The core idea is that running large-scale production systems is itself a software engineering problem — automation, observability, capacity planning, and incident response should be approached with the same rigor as product development.
Key SRE concepts include Service Level Objectives (SLOs), error budgets, blameless postmortems, toil reduction, and the distinction between planned work and reactive work. Most large technology companies now have SRE or equivalent functions, even if the title varies (DevOps, Platform Engineering, Production Engineering, Infrastructure).
For AI-deployed systems, SRE practices are actively being extended to cover agent-specific reliability concerns: what counts as "down" when a model is degraded but not failing, how to measure reliability for probabilistic outputs, how to set SLOs for agent task completion rates. This is an actively evolving area where standard practices have not yet crystallized.