Chief AI Officer Interview Questions That Reveal Something
Most AI leadership interviews test vocabulary. Vocabulary is the cheapest thing a candidate can acquire.

Erin Moore
Fractional Chief AI Officer
Ask about things that went wrong, things they stopped, and things they decided not to do. Candidates rehearse their successes and almost nobody rehearses their cancellations — which is where you find out whether someone has actually run a portfolio or has only presented one.
Twelve questions, in four groups, with what a strong answer sounds like.
Judgement
"Tell me about an AI project you killed. Who was unhappy?" The best answer names a specific project, a specific person who was unhappy, and what the candidate did about it. Anyone who has run AI at scale has killed things; a candidate with no cancellations has either not had authority or is not telling you about them. The reasoning is in how to kill an AI project.
"What did you believe about AI two years ago that you no longer believe?" Tests whether they update. Weak answers are about vendors; strong answers are about their own method.
"Where would you tell us not to use AI?" A candidate with no answer is selling. Strong answers are specific to your business and arrive fast, which means they were thinking about it during the earlier conversation.
Method
"Walk me through how you would spend your first ninety days here." Listen for whether it starts with an inventory and interviews or with a technology plan. The better shape is in the Chief AI Officer's first 90 days.
"How do you decide which use cases to fund?" You want an actual framework, not "business value". Listen for whether reversibility appears anywhere — per AI use case prioritization, it is the axis that separates people who have shipped from people who have planned.
"How do you know whether a deployed system is still working?" Tests whether they think past launch. Weak answers describe a dashboard. Strong answers describe an evaluation set, a re-run cadence, and who looks at it.
"A business unit bought a tool without telling you. What do you do?" Strong answers start by asking why they went around the process. Candidates who lead with enforcement will reproduce the conditions that create shadow AI.
Standing
"What did you need authority over in your last role, and did you have it?" Reveals whether they understand that this role fails structurally more often than personally. Follow with what they would need here — a candidate who does not raise decision rights has not thought hard about the job, per who owns AI governance.
"Describe a time you disagreed with a CEO." You are hiring independent judgement. Someone who cannot produce an example will not produce one for you either.
"Who should this role report to, and why?" Tests whether they will negotiate for the conditions they need. The trade-offs are in Chief AI Officer reporting structure.
Substance
"Explain retrieval-augmented generation to our CFO." Tests communication, not knowledge. Any credible candidate knows what it is; far fewer can explain it without jargon to someone holding the budget.
"What would you need to see to tell the board we should slow down?" Strong answers include specific thresholds. This one question does more than any technical screen to reveal whether the person will tell you an inconvenient truth in month eight.
Three answers that should end the process
A guaranteed ROI figure before they have seen your data. A recommendation of a specific vendor before understanding the use case. And an inability to describe anything that has gone wrong on their watch — which either means no real ownership, or a candidate who will manage your perception rather than your risk.
For an external standard to hold the conversation against, NIST's AI Risk Management Framework and its playbook are what a serious candidate will already be familiar with, and Stanford HAI's AI Index is a reasonable check on whether their sense of the field is current or from three years ago.
Frequently asked questions
What should you ask a Chief AI Officer candidate? Ask about failures, cancellations and decisions not to act. Candidates rehearse successes; almost nobody rehearses the projects they killed, which is where genuine portfolio ownership shows.
What is the single best interview question for this role? "Tell me about an AI project you killed, and who was unhappy." Anyone who has run AI with real authority has cancelled things, and the answer reveals both judgement and standing.
What answers should disqualify a candidate? A guaranteed ROI figure before seeing your data, a vendor recommendation before understanding the use case, and an inability to name anything that went wrong on their watch.
Should we test technical depth? Some, but communication matters more. Ask them to explain retrieval-augmented generation to your CFO — most credible candidates know the concept, and far fewer can explain it clearly to the person holding the budget.
Further reading
Tagged with:
Ready to Automate Your Business?
Let's discuss how AI automation can deliver measurable ROI for your organization in 90 days or sooner.