Organize
Who owns AI, and how does adoption spread?
Ownership of AI is set by structure, not by title. A CAIO with no authority over use case approval, vendor procurement or model standards produces the appearance of governance without the substance, and a structure chosen by default stops fitting once the portfolio scales. The first half of this discipline covers getting ownership right: what the CAIO mandate has to include, which structural model fits which stage, how decision rights name one accountable owner per decision, and what keeps CIO, CDO, CISO, Legal, Finance and CHRO coordinating instead of each running AI on its own timeline.
Adoption then spreads through people, and it is not self-managing. Roles change composition as agents absorb routine work. Middle managers decide which tools their teams treat as safe and how performance is measured. Experts decide whether their judgment gets captured in forms agents can use. Every task needs a deliberate choice about how humans and agents share the work.
The thread through both halves is the same: name the owner, make the trade-off explicit, and put a workforce plan behind every deployment.
The CAIO Mandate
What the CAIO role has to own, where it should report, and why an advisory CAIO produces governance in appearance only.
Structural Models
Centralized CoE, hub-and-spoke or federated: when each model fits, how each breaks, and how to move between them.
Decision Rights
A RACI-style matrix that names one accountable owner for each AI decision, from use case approval to agent action boundaries.
Cross-Functional Coordination
How CIO, CDO, CISO, Legal, Finance and CHRO actually align: steering committee, shared OKRs, joint funding and one integrated roadmap.
Role Evolution
Which roles shrink, which grow, which emerge, and why workforce planning belongs inside the AI roadmap rather than beside it.
The Middle Management Gap
Why AI programs stall in the middle layer, what managers actually control, and how to turn passive resistance into championship.
Knowledge Architecture
Why agents fail at the boundary of what is written down, and the four components that turn tribal knowledge into something agents can use.
Human-Agent Collaboration
Five collaboration patterns from copilot to full autonomy, how to match each to the task, and why full autonomy is an end state, not a start.
Decision Records
Five ADR templates for the choices that shape an AI program: operating model, governance build or buy, agent authorization, measurement, and shadow AI response.