Everyone Wants An AI Executive, Yet Nobody Can Define One

This voice experience is generated by AI. Learn more.This voice experience is generated by AI. Learn more.Shawn Cole is the president and cofounder of Cowen Partners, with 20+ years of experience building and growing successful businesses.

getty​A prospective client came to my business, an executive search firm, not long ago asking for a chief executive focused on AI. That was the requirement as stated, but when asked why, no further detail followed because, in the client’s mind, none was needed.

I’m seeing that request show up across industries and functions now, and the pattern underneath it is remarkably consistent. Clients know the competition for technical talent is fierce. They know they are behind where they want to be. What they don’t know is what “AI expertise” would actually need to look like in the person sitting across the table.

To understand this evolving role, we ran a search of Securities and Exchange Commission full-text filings for “chief AI officer” and similar formalized titles across recent proxy statements and current reports. What we found is a small number of formally titled positions relative to how often boards ask for this kind of expertise in a search. That gap is itself worth noting. The demand for AI leadership is running well ahead of any shared definition of what it means, and almost nobody is asking the follow-up question that would close that gap.

When an AI requirement shows up in a search, we ask the client one question: What would a candidate need to have actually done for this requirement to count as satisfied?

That question almost always produces a pause. Most clients haven’t had to translate “AI expertise” into a specific action or outcome before because nobody has asked them to. The pause is where the real specification gets written, as what comes out the other side is usually a lot narrower and a lot more useful than what went in.

Over enough of these conversations, a pattern emerged. We think about AI expertise in four levels, and I recommend treating them as levels of evidence a candidate can demonstrate, rather than a ranking of better or worse executives.

Level One, AI Curious: The executive uses the tools personally and follows the space closely enough to hold an informed conversation. There’s no operational claim attached to this level, just genuine awareness.

Level Two, AI Proficient: The executive has changed how they personally work because of these systems. They use AI daily in their own decision-making and can describe specifically what they stopped doing manually as a result.

Level Three, AI Applied: The executive has deployed AI into a function and owns the outcome of that deployment. They can describe what they implemented, what it cost, what broke along the way and what it actually returned.

Level Four, AI Fluent: The executive understands the economics underneath the tools themselves, including model classes, where inference costs sit, build versus buy decisions, what the data requirements really are and which problems aren’t worth solving this way at all.

The right level depends entirely on what the role in your company will actually have to decide. For example, a chief financial officer approving AI capital spending needs level four because the job is judging whether the projected return is real. Alternately, a chief executive typically needs level three across the board, with genuine level four judgment reserved for one or two decisions that actually matter to the company’s capital allocation.

Every company is using AI differently right now, which means the specification has to get written against your company’s actual decisions, rather than against the category of AI expertise in general.

I’ve found asking a few questions directly in an interview tend to separate these levels cleanly.

1. What have you personally stopped doing manually in the past 18 months? This separates level one from level two, and it’s difficult to answer convincingly without real adoption behind it.

2. Describe something you deployed, what it cost and what it actually returned. This separates level two from level three. Vague answers here are the single most common signal that a candidate is describing familiarity rather than ownership.

3. What AI project did you decide not to pursue, and why? This separates level three from level four, since judgment about what not to build requires understanding the underlying economics well enough to walk away from something that sounds appealing.

4. How would you know within a year whether this investment actually worked? This tests whether the candidate thinks in terms of returns or simply in terms of adoption.

For a commercial HVAC client we worked with recently, this process resolved the vague request into something specific. The role needed an executive who used these systems personally in their own work and could connect AI-enabled analysis directly to demand forecasting. That’s a real specification. “AI expertise” on its own is not.

The more valuable half of this exercise happens before the search even opens. Write down the three the executive will actually have to make about AI in their first two years. The level of expertise required falls out of those three decisions on its own, and the specification writes itself from there.

The requirement for AI leadership is likely going to keep arriving in searches across every function and every industry. What’s actually missing is a shared definition of what satisfies it, one specific enough that a board can write it into a job description and a candidate can be evaluated against it honestly. Without that definition, every search defaults to the same vague word, and every candidate gets measured by a different, unspoken standard depending on who’s in the room.

I’m offering these four levels as our contribution to that shared definition. A board or a chief executive who can name the level a role actually requires can walk into a search already ahead of most of the market it’s competing in for this kind of talent, because the other boards are still writing job descriptions around a word instead of a standard.​

Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?