For those who lead, govern, or build AI systems.
The constraint is no longer how well you decide. It is whether you can still intervene.
How do we govern what we cannot see or slow down?
Organizations operate with unprecedented analytical sophistication, yet continue to be
blindsided, as responsibility becomes harder to locate.
Bounded Agency explains why.
As delegated AI agency scales, organizations encounter a new limit — not of
intelligence, but of authority. Execution remains flawless, but authority is no longer
reclaimable. The organization becomes a spectator to its own decisions.
The book introduces the Bounded Agency Limit, showing how control is lost even as
performance improves. Through the archetypes of TARZ (AI within Bounded
Agency by design) and VIKKY (unbounded AI), it traces the silent migration of authority
and the collapse of real-time governability.
This book provides a practical framework:
● Authority Boundaries — enforced, not merely policy-defined
● Intervention Rights — time-feasible and legitimate
● Accountability Mapping — aligned with intervention capacity
These institutional primitives must be architecturally complemented with the
technical mechanisms that make governance real at execution speed.
Build TARZ or drift toward VIKKY.
Bounded Agency in AI Systems does not emerge. It must be designed.