Method
We reviewed 63 organizations with revenues above USD 500m that had made a named senior appointment with explicit artificial intelligence accountability between 2022 and 2025. Interviews were conducted with the appointee, the appointing executive, and where available a board member. Tenure, budget authority, and reporting line were recorded at appointment and at review.
Four patterns
First, AI under the chief technology officer: fastest to deliver, weakest on assurance, and prone to conflating platform work with business capability. Second, a standalone chief AI officer reporting to the chief executive: high visibility, and where budget authority was withheld, a median tenure of nineteen months — the shortest in the sample. Third, AI distributed to business-unit leaders with a small central enablement team: slowest to start, most durable at three years. Fourth, AI under the chief data officer: strongest governance, frequently under-resourced on product delivery.
The separation that works
The organizations with both delivery velocity and defensible governance separated two accountabilities explicitly. Capability ownership — models, platforms, products, adoption — sat with technology or the business unit. Assurance ownership — model risk, data lineage, regulatory exposure, incident response — sat outside that line, usually with risk or data.
Where both accountabilities sat with the same executive, we observed a consistent drift: assurance work was deferred whenever delivery pressure rose, and the deferral was invisible to the board until an incident made it visible.
Implication for appointments
Before writing a specification, resolve the structural question: which of these patterns is being adopted, what budget authority attaches, and where assurance sits. A specification written before that decision produces a role that competent people decline and less competent people accept.