The Human in the System
The book has argued that agentic reliability engineering is the natural successor to the SRE discipline, not a departure from it. The argument is worth restating in one tight pass before the book closes.
What changes is where the work happens. The reliability work that used to be done by a human reading a dashboard now happens in the Observability Agent’s enrichment pipeline. The reliability work that used to be done by a human deciding whether to roll back now happens in the Release Agent’s CRS computation, with the orchestrator reconciling against intent. The reliability work that used to be done by a human approving a capacity expansion now happens in the Capacity Agent’s autonomous T2 authority. The work has not disappeared; it has moved. The pre-agentic team did the work directly; the agentic team designs the substrate that does the work, governs the substrate’s behavior, and reviews the substrate’s calibration over time.
What changes is what gets automated. The repetitive, the well understood, the high-volume, and the easily reversible: these become agent action classes operating at T2 or T3 authority. The novel, the high-stakes, the irreversible, the politically consequential: these stay with humans, often with explicit authority tier 0 or tier 1 binding. The boundary between the automated and the human is itself an engineering artifact, reviewed quarterly, and evolved deliberately as evidence accumulates. Pre-agentic engineering knew where its automation ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access