\cleardoublepage \part{Managing Machines} \setbookpart{Managing Machines}

Part IV: Managing Machines

The first three parts looked at AI that manages people. This part looks in the other direction. In knowledge work the more common experience of AI is not being managed by it but managing it: assigning tasks to agents, reviewing what they produce, combining their output with the work of colleagues and answering for the result. Microsoft's Work Trend Index describes every employee becoming an "agent boss". The phrase is the vendor's, and this book quotes it rather than adopts it, but the development it describes is observable, and it changes what a manager is.

Managing agents is not managing people. An agent has no career, no health and no voice to protect, so M4 Development and M5 Care do not transfer, while M1 Direction and M2 Evaluation transfer in a changed form: the manager writes a mandate rather than gives an instruction, and verifies output rather than appraises a colleague. What does transfer in full is accountability. When an agent orchestrates other agents, and when people and agents work in the same team under the same manager, the chain of accountability must still end with a named person, and the decision rights along the chain must be written down.

Chapters in this part:

  • Chapter 16, The Agent Boss: When the Team Is Software, examines what a person does when the team is software, introduces the mandate manifest for the agents a manager directs, and shows which management functions transfer to the management of agents and which do not.
  • Chapter 17, Orchestrators: AI Managing AI Agents and the Chain of Accountability, examines orchestrator agents that assign work to other agents, where the chain of accountability ends, and how it is recorded.
  • Chapter 18, The Mixed Team: People and Agents Under the Same Manager, examines teams in which people and agents share a queue, how work is allocated fairly between them, and how a manager evaluates people whose output is entangled with agents' output.
  • Chapter 19, Skill, Judgement and Who Becomes a Manager When AI Does the Managing, examines what happens to the junior management work through which people used to learn to lead, and how organisations keep a route into management.
  • Chapter 20, Decision Rights in the Management Chain: Who May Direct, Evaluate, Reward and Dismiss, sets out the written decision rights for each management function along a chain that includes people, agents and orchestrators.

Management functions in this part. Chapters 16 and 17 concern chiefly M1 Direction and M2 Evaluation as they apply to agents. Chapter 18 adds M3 Consequence, because the evaluation of a person in a mixed team can be distorted by the agents beside them. Chapter 19 concerns M4 Development, for the manager as well as the managed. Chapter 20 covers all five and turns them into decision rights. Every chapter names the functions it concerns in its opening and its Summary.

The thread from the series and from Part III. The Transformed Security Architect concluded that an agent team works under a written mandate manifest at an autonomy level matched to reversibility, and that the person who conducts it remains the accountable verifier of everything it proposes (The Transformed Security Architect, Chapter 13). That was Shift 4 of that book, from sole author to conductor of an AI team, applied to one role. This part applies it to every line manager. Part III showed that the harms of the AI boss concentrate where accountability is unclear; Part IV shows how to keep it clear when the chain includes machines on both sides of the manager.

What you will learn in this part:

  • Which management functions transfer to the management of agents, and which remain with people
  • How to write a mandate manifest for the agents a manager directs
  • Where the chain of accountability ends when agents orchestrate agents
  • How to evaluate people fairly in a team they share with agents
  • How to keep a route into management when the junior management work is automated
  • How to write decision rights for direction, evaluation, reward and dismissal

Part V turns to the law: how Europe has answered the AI boss, and what its answer requires of employers.