The call I get most often starts the same way. A CEO somewhere between $20M and $500M in revenue has spent real money on AI advice and has a document to show for it. Sometimes it is a very good document: a landscape scan, a maturity model, a dozen use cases scored on effort and impact, a recommendation to start with three of them. Six months later that CEO can still tell me what the document said. What they cannot tell me is what changed in the business. Nobody was lied to and the analysis was not wrong. It just never turned into decisions.
That gap is not a quality problem. It is a role problem. An AI consultant and an AI executive are two different jobs, sold at similar price points, described with overlapping language, and confused constantly. The difference is not seniority, and it is not how smart the person in the room is. Some of the sharpest AI thinkers I know are consultants. The difference is who owns the outcome after the engagement ends, and that one distinction changes what you should buy, what you should pay, and what you are allowed to expect.
A consultant sells you analysis and a recommendation. That is the product, and when it is done well it is genuinely valuable. You are buying pattern recognition across many companies, capacity your team does not have, a market scan you would take four months to run yourself, and an outside voice with enough standing to say the thing your own people have been too political to say out loud. If you need to know what the vendor landscape looks like, whether your peers are further ahead than you, or which of nine platforms survives a structured evaluation, hire a consultant. That work is real, it is hard, and it is worth paying for.
What you are not buying is authority. A consultant does not sit in your operating rhythm, cannot tell your VP of Operations that her team is going to change how it works next quarter, and is not in the room in month seven when the vendor slips and someone has to decide whether to push the date or cut scope. The engagement has an end date and the recommendation is the deliverable. When the recommendation lands on a desk that has no time to act on it, nothing happens, and that is not the consultant's failure. It was never their job.
An executive sells you decisions and consequences. The product is not a document. It is a sequence of judgment calls made inside your company, in your context, with your constraints, by someone who has to live with them.
In practice that means four things a consultant structurally cannot do. An executive says no on your behalf, which is most of the job: killing the three projects that would have burned a year, telling the board that the flashy pilot is a distraction. An executive holds vendors accountable, because they will still be there when the invoice arrives and the milestone did not. An executive makes the sequencing calls, deciding what happens in which order given the team you actually have rather than the team the plan assumes. And an executive tells you the truth about your own organization, including the parts about your people and, occasionally, about you.
A consultant is accountable for the recommendation. An executive is accountable for what happens after it.
That is the entire distinction, and every practical difference falls out of it. It explains why the consultant's engagement ends when the deck is delivered and the executive's does not. It explains why a consultant can be brilliant and change nothing, and why an executive can be less credentialed and change quite a lot. It also explains the pricing logic. You pay a consultant for their time and their pattern library. You pay an executive to carry risk on your behalf.
When a company asks me which they should hire, I do not ask about budget first. I ask three questions, and the answers usually settle it in about ten minutes.
The first: what happens to the artifact when the engagement ends? If the honest answer is that someone on your team will pick up the recommendation and run it, and you can name that person and point at the room in their calendar where it happens, a consultant is the right buy. If the honest answer is that it goes into a shared drive and gets referenced in a board meeting in November, you do not have a capacity problem that analysis will fix. You have an ownership gap, and buying more analysis makes it worse.
The second: who is in the room when the vendor pushes back? Every AI project has the moment where the partner says the integration is harder than scoped and the date is moving. Whoever is sitting on your side of that table making the call is your actual AI leader, whatever their title is. If that person is your CFO in a meeting they did not want to be in, working from a scoring matrix somebody else built, you are running the wave without a leader.
The third: is the decision in front of you a "what" or a "how"? What to do is a consulting question, and a good one. Which workflow, which vendor, which sequence, all of that responds well to structured analysis. How to actually make it happen inside a company that has a busy quarter, a skeptical operations lead, and a data environment nobody wants to open, is an execution question. Execution questions do not respond to better analysis. They respond to someone with the authority to move things.
Almost nobody buys the wrong role on purpose. The middle market ends up there because of a structural gap, and it is worth naming.
A Fortune 500 company runs both without thinking about it. There is a Chief Digital or Chief Data or now Chief AI Officer on the org chart who owns the outcome, and there is a tier-one consultancy engaged underneath them for surge capacity and analysis. The roles are separate, and everyone knows which one they are talking to. A $60M manufacturer has neither. There is no executive whose job is AI, and the consulting firms that would do serious work at that scale are priced for companies ten times larger. So the company buys the thing that is available and affordable, which is analysis, and hopes it will behave like leadership. It will not, and six months later the CEO has that very good document and no change in the business.
I wrote about the reflex on the other side of this in why your first AI hire shouldn't be a Chief AI Officer. The failure mode there is the mirror image: a company that correctly senses it needs an owner, and reaches for a full-time executive search it cannot yet justify or fill. Both mistakes come from the same place, which is a real need for ownership and no obvious way to buy it at your size.
A fractional AI executive is not a consultant with a friendlier invoice. The test is the same one above: does this person own outcomes, or do they own a deliverable? A fractional executive sits in your leadership meetings, has decision rights you have explicitly given them, carries a number, argues with your vendors, and is still there in month seven. They just do it two days a week instead of five, which is the correct amount of AI leadership for a company that has one or two real bets running rather than forty.
The reason the shape works at this size is not the cost saving, though the cost saving is real. It is that senior operating judgment is the scarce input, and it is lumpy. The hard parts of an AI program are a handful of decisions, made at the right moments, by someone who has watched this movie before. The rest is execution your team can do. Paying full-time comp to have that judgment sitting idle between decisions is a worse deal than most people think, and for many middle-market companies it is not a deal that is available at all.
None of this is an argument against consultants. The pattern I like best runs both, in order. Start with a tight, time-boxed piece of analysis to establish the map: which workflows are worth the money, what the data situation actually is, what a real sequence looks like. That is consulting work and it should be scoped, priced, and finished. Then put someone with ownership behind it to make the decisions the map implies.
What I would not do is buy the analysis and stop, then wonder in the fall why nothing moved. And I would not put an owner in place with no map, because they will spend their first two months building one anyway, at a higher rate. The sequence matters more than the labels, and the failure mode to watch for is buying the first half of it twice.
Answer the three questions honestly. If you have a named owner with the time and standing to act, and what you are missing is analysis, hire a consultant and hold them to a sharp scope. If you have plenty of opinion in the building and nobody who owns the outcome, more analysis is the expensive way to postpone the decision, and what you need is an executive, fractional or otherwise.
More on this: the fractional AI executive page describes what the embedded version looks like week to week, and the AI Opportunity Sprint is the two-week mapping exercise that comes first. Why middle-market AI strategies fail covers what happens when the ownership question goes unanswered, and don't build an AI Center of Excellence is the same argument at the level of the whole org. If you want a second opinion on which of the two you are actually shopping for, a twenty-minute call is the place to start.
Twenty-minute scoping call. No slide deck, no pitch. We talk about where you are and whether a Sprint or a Fractional engagement fits.
Book a scoping call