AI is becoming an increasingly important actor in decision-making. In this new era, what does it mean for an organization to remain “governed by humans”?

This book confronts that fundamental question through a rigorous theoretical framework.

As more and more tasks are delegated to AI, consequential judgments and choices about values still remain in human hands. But does human oversight actually work? Can humans oversee everything? Where does delegation end—and where must it stop?

This book reframes that boundary as something that can be defined and measured.

At its core is the concept of Oversight Capacity. The book explains the conditions and patterns under which human oversight breaks down, and identifies which responsibilities ultimately cannot be delegated. In environments where AI operates autonomously, it redefines the human role through three lenses: authority architecture, attention capital, and verification independence.

What This Book Reveals

Oversight is not an “attitude” but an architecture of authority, and the appropriate position of human oversight is conditional.

Oversight capacity is a finite resource, and failure can occur not through “refusal” but through acceptance under overload.

The uniquely human contribution is not necessarily “better judgment,” but independence of verification.

There is a boundary between domains in which different forms of oversight can scale and those in which they cannot.

At the end of delegation lies a Non-Delegable Residual—a set of responsibilities that must ultimately remain with humans.

The theory does not remain abstract.

The book proposes observable indicators for real-world practice, together with a declaration protocol that organizations can use to examine their own governance structures. It provides a foundation for governance design and accountability in the age of AI.

AI expands capability. But humans still choose.

This book returns to that premise and asks what governance must become in an AI-driven world.

Who This Book Is For

Executives, board members, CFOs, CHROs, legal and risk leaders, AI and data leaders, policymakers, researchers, graduate students, and anyone concerned with the future of governance in the age of AI.

Length: approx. 54,000 words

Why Human Governance Breaks

Comentarios