This book covers the design, operation, and governance of AI agents built on Large Language Models. It works through the architecture of an agent, the model at its core, the tools it calls, the memory it keeps, and the flow that coordinates them, and then turns to the organizational layer: how agents are integrated into existing systems, what they cost, which business processes suit them, and what policies they require. A third part treats operations, covering the orchestration of several agents, runtime governance, maintenance over time, and security against attacks that target the model itself.
The treatment is organized around one framework that connects model behavior to architectural components, orchestration logic, governance mechanisms, and organizational context. Executable Python examples anchor every mechanism and are written to be vendor-neutral, so they outlast the tooling they demonstrate.
The book grew out of research, teaching, and applied projects at the Institute for Process Management and Digital Transformation at FH Münster. A recurring pattern there shaped it: obtaining useful results from an AI agent is easy, while designing a robust, auditable, and maintainable system around one is not.
