\cleardoublepage \part{Foundations} \setbookpart{Foundations}

Part I: Foundations

Artificial intelligence changes what an organisation must govern. The systems in this book behave probabilistically, derive their most important component from data, depend on third-party models whose training the organisation cannot inspect, and operate under a European legal framework that assigns obligations to roles many organisations have not yet recognised that they hold. Every later part assumes that the reader can see those four facts, and this part supplies the footing: a shared vocabulary, a map of the law and a small set of governance principles.

Part I sets out what AI governance is, why it belongs in version control rather than in slide decks, and how this book follows Architecture as Code in The Transformed Architecture. It introduces Nordvik Region and Hansa Mutual, whose decisions run through every chapter. It then distinguishes the usage modes of large language models, from ambient AI to autonomous agents, because each mode needs different controls. It maps the AI Act and the surrounding legislation, sets out the guiding principles, and closes with readiness, because the most expensive AI failures occur when budget is committed before data, platform, governance and skills can carry the weight.

What you will learn in this part:

  • What AI governance is, and why it is most effective when expressed as code
  • The six usage modes of large language models, why risks and controls differ between them, and how to classify an AI estate by mode
  • How the AI Act's roles, risk classes and timeline translate into governance requirements, and how it interacts with the GDPR, DORA, NIS 2, the Data Act and the Cyber Resilience Act
  • The eight principles that guide every later governance decision
  • How to assess readiness across data, platform, governance, skills and culture

Everything after Part I depends on three of its distinctions: which legal role the organisation holds for a system, which usage mode the system belongs to, and which principle a given control serves.