The AI Gap · Tool

The VP Org Design Check

Twelve questions to answer before you approve an AI, analytics or data leadership req — not after the shortlist arrives.

Most failed senior data and AI hires are not screening failures. They are structure failures locked in before the first interview. The role was defined against the wrong constraint, the reporting line was left to the offer stage, and the decision rights were never written down.

This takes about ninety minutes with the hiring manager and the two peers the role has to work with. It is answered before the job description is written.

Section 1

The constraint

93.2% of data and AI leaders name culture, change and organization as the greatest obstacle to AI adoption. 6.8% name technology. Most senior specs are written against the 6.8%.

1. What is the actual constraint this role removes?

Write it as a sentence with a subject: "Business unit leaders will not change their process to use the model." Not "we need stronger ML capability." If the constraint is genuinely technical, prove it before assuming it.

2. What has already been tried against that constraint, and why did it fail?

If nothing has been tried, the role is being hired to discover the problem. Level and price it accordingly, and expect an eighteen-month clock rather than twelve.

3. What does year one look like in outcomes the business already tracks?

Deployment count is not an outcome. Adoption rate, cycle time, cost per case, retained revenue are. Across the Fortune 1000 benchmark, 39.1% now run AI in production at scale while 54% report high or significant business value. Deployment stopped being the differentiator.

4. If this role fully succeeds, what does someone else stop doing?

Every real mandate takes something from someone. If you cannot name whose scope changes, the role has authority on paper only.

Section 2

The structure

Where AI leadership reports has no market consensus: technology 33.9%, the data function 30.4%, business 26.8%, transformation 8.9%. There is no default to inherit quietly.

5. Where does the role report, and who was in the room when that was decided?

Decide before the search. A reporting line negotiated at the offer stage is one the rest of the leadership team never agreed to.

6. Does the role own standards, delivery, or both?

Standards without delivery produces a governance function nobody consults. Delivery without standards produces four teams building the same thing. Choose knowingly.

7. What can this person say no to, and who can overrule it?

Name the person. If the answer is "they'd escalate," the role has influence rather than authority — and the candidate profile should change to match.

8. Who funds the platform this role depends on?

If it sits in another function's budget, year one is a negotiation rather than a build. That is a legitimate design, but hire someone who has done it before.

9. What happens to this structure if the sponsoring executive leaves?

CDO tenure past five years has risen to 25.8%, which still leaves three quarters of the seat turning over faster. A structure that survives only under one sponsor is not a structure.

Section 3

The bench

Data and analytics postings have been falling while the share mentioning AI climbs past 45%. The mid-level work that used to train senior people is the work being absorbed first.

10. Where does this role's successor come from?

If the answer is the external market, budget for that search now and say so out loud.

11. What do the junior seats under this role actually learn?

If the answer is throughput — dashboards, ad hoc queries, reconciliation — you are staffing a layer that is disappearing and training nobody. Junior seats have to be designed to teach problem framing.

12. Who owns adoption after launch?

If it is the team that built the thing, adoption is a side project. If it is nobody, you have your answer about the value gap.

Put three of these in the loop

Scored, with a rubric, weighted equally with the technical questions.

Constraint

Describe a process you changed that people did not want changed. What did you give up to get it?

Structure

Walk me through a reporting line that made your job harder. What did you renegotiate, and when?

Bench

Who did you promote into a role you used to hold, and what did you build to make that possible?

Unscored questions lose to scored questions in every debrief. That single mechanic is how a spec written for the 6.8% survives a panel that knew better.

Sources

2026 AI & Data Leadership Executive Benchmark Survey, Data & AI Leadership Exchange (~110 Fortune 1000 and global firms, 96% C-level). McKinsey, The State of AI, November 2025. Indeed Hiring Lab, 2026 US labour market reporting.

This is from The AI Gap, a monthly newsletter on the distance between what the evidence shows and what organizations do.

Subscribe on LinkedIn →