I am asked about once a month whether a company should appoint a Chief AI Officer. My answer disappoints whoever is asking: it depends what you are trying to avoid. If the goal is to have someone to point at when a director asks about AI, the title will do that. If the goal is to have AI actually change how the business operates, a new box on the org chart is the slowest possible route.
Why the title backfires
AI is not a function. It is a capability that has to live inside the functions that already own the customers, the products and the risk. The moment one executive owns AI, everyone else is relieved of it. Marketing waits for the AI team. Operations waits for the AI team. The AI team, staffed with people who know models but not the business, spends a year building things nobody asked for. I have watched this film three times.
The questions that do the work
What boards need instead is a short, repeatable set of questions that they ask every leader, every quarter, until the answers stop being embarrassing:
- Where in your P&L is AI already changing a number, and by how much?
- Which decision in your area would you never let a model make, and why?
- What did we learn from the last AI initiative that did not work?
- Who owns the data your models depend on, and would they say so?
A good question, asked by the board every quarter, moves an organisation further than any title ever will.
From a board offsite, spring 2026
When a dedicated role does make sense
There are exceptions. A regulated business facing hard governance deadlines may need a single accountable executive for a period. A company whose product is a model needs an AI leader the way a bank needs a CRO. In both cases the role should have a sunset clause and a clear job: build the capability into the line, then hand it over. If the role still exists in five years, it has failed.
