\cleardoublepage \part{The Craft} \setbookpart{The Craft}

Part II: The Craft

Operations is craft before it is AI. Part II treats each function in turn: what is observed and how, how a fault is diagnosed, how change is made and made safely, how service is restored, and how the estate and the learning are sustained. Each chapter first describes the practice as it stands, then what AI adds, then what can go wrong.

The weight of the book lies here. The chapters on restoring (Chapter 10) stage a debate on whether AI should remediate production on its own.

Chapters in this part:

  • Chapter 5, Observe, treats telemetry, objectives and error budgets without AI first (O1 Observe).
  • Chapter 6, Observe with AI, treats correlation, anomaly detection and alert noise with AI (O1 Observe, O2 Diagnose).
  • Chapter 7, Diagnose, treats triage, root-cause analysis and the incident assistant (O2 Diagnose).
  • Chapter 8, Change, treats releases, infrastructure as code and AI-authored change (O3 Change).
  • Chapter 9, Change Safely, treats progressive delivery, freezes and blast radius (O3 Change, O4 Restore).
  • Chapter 10, Restore, treats runbooks, auto-remediation and recovery, and stages a debate (O4 Restore).
  • Chapter 11, Sustain, treats capacity, cost, energy and lifecycle (O5 Sustain).
  • Chapter 12, Sustain, treats problem management, post-incident learning and knowledge (O5 Sustain, O2 Diagnose).

Functions in this part. Part II covers all five functions: O1 Observe in Chapters 5 and 6, O2 Diagnose in Chapters 6, 7 and 12, O3 Change in Chapters 8 and 9, O4 Restore in Chapters 9 and 10, O5 Sustain in Chapters 11 and 12.

The thread from the series. Observability and testing as code are those of Architecture as Code; evidence is the subject of AI Governance, Chapter 11; detection as a record and closure by class are those of AI Blue, Chapters 6 and 7; the independence of the re-test is that of AI Purple, Chapter 11.

What you will learn in this part:

  • How a service level objective is written and spent
  • What AI adds to detection and what it can suppress
  • How an incident assistant is used without anchoring
  • How AI-authored change is released safely
  • Which remediation actions a machine may take

Part III turns to the systems that operations must now run: AI systems themselves.