Fractional AI officer, or fractional CMO?

“Fractional AI Officer” appeared as a job title roughly eighteen months ago and is now on a great many LinkedIn profiles, including people who were describing themselves as something else last year. That makes it reasonable to ask whether it is a real role or a relabelled one.

It is sometimes real. Here is how to tell which you are being offered, and which you actually need — which are different questions.

What each role actually owns

Forget the titles for a moment and look at what sits on the desk.

A fractional CMO owns demand. Positioning, pricing, channels, the funnel, the team that executes it, and the number at the end. The question they answer is “why are we not growing, and what changes that.” They are accountable for a commercial outcome.

A fractional AI officer owns capability. Which processes should use AI, which tools, what the data and governance look like, what the staff need to know, and what the organisation is not allowed to automate. The question they answer is “how does this company operate this technology safely and usefully.” They are accountable for an operational outcome.

Those are genuinely different jobs. The overlap is real but narrower than the titles suggest — it is mostly the part where AI changes how customers find you, which is one slice of each role.

How to tell a real AI officer from a relabelled one

The title is new enough that it carries no signal. The work does. Four questions separate them quickly.

“What would you tell us not to automate?” The most useful question in the set. Someone who has done this work has a ready answer, usually involving judgment calls, regulated decisions, and anything where being wrong is expensive and hard to detect. Someone who cannot name limits is selling tools.

“Walk me through a process you mapped before recommending anything.” Real engagements start with understanding how work currently happens. If the answer jumps straight to a tool stack, the diagnosis step is missing.

“How would you measure whether this worked?” Good answers are operational and boring: cycle time, error rate, hours returned, adoption. Weak answers are about being more innovative.

“What happens when the model changes?” The field moves fast enough that anything built on one vendor's current behaviour needs a plan for when that behaviour shifts. Someone who has operated through a model deprecation will say so.

Which one you probably need

Honest answer for most businesses reading this: the CMO, or neither.

If your problem is that not enough customers are coming, that is a demand problem. It is almost never solved by an AI capability programme. You need positioning, a clear offer and working channels. A fractional CMO. Adding AI tooling to a business with unclear positioning produces the same confusion, faster.

If your problem is that the work is drowning you — the operation cannot keep up, errors are creeping in, people are doing things software should do — that is a capability problem, and an AI officer is the right shape. This is where the role earns its keep.

If people across your company are already using AI tools nobody approved, on customer data nobody reviewed, you have a governance problem that is currently invisible. That is the clearest case for the role, and most companies in it do not know they are.

If you are under about ten people, you likely need neither as a standing role. You need a few days of help making specific decisions, then to get on with it.

On buying both

Occasionally a company genuinely needs both, and occasionally one person can do both — I do, for a narrow band of companies where marketing is the dominant function and the AI question is mostly about how customers find them.

But be suspicious of anyone who offers both without qualification. They are different skill sets and the honest version of “I do both” comes with a boundary attached. Mine: I am the right person when the AI question is about visibility, positioning and how customers encounter you. I am the wrong person for enterprise data governance or ML engineering, and I will say so rather than learn on your budget.

If you are talking to someone whose offering has no edges, you have not found the edges yet.

Common questions

What is the difference between a fractional AI officer and a fractional CMO?
A fractional CMO owns demand — positioning, pricing, channels, the funnel and the revenue number — and is accountable for a commercial outcome. A fractional AI officer owns capability: which processes should use AI, which tools, what the data and governance look like, what staff need to know and what should not be automated. They are accountable for an operational outcome. The overlap is mostly the part where AI changes how customers discover a business.

Which do I need for my business?
If the problem is that not enough customers are arriving, that is a demand problem and a fractional CMO is the right role; AI tooling will not fix unclear positioning. If the problem is that operations cannot keep up, errors are creeping in, or staff are doing work software should do, that is a capability problem and an AI officer fits. If people are already using AI tools nobody approved on data nobody reviewed, that is a governance problem and the clearest case for the role. Under roughly ten people, usually neither as a standing role.

How do I tell a real fractional AI officer from a relabelled consultant?
Ask what they would tell you not to automate — experienced practitioners have a ready answer involving judgment calls, regulated decisions and situations where errors are costly and hard to detect. Ask them to describe a process they mapped before recommending any tool. Ask how they would measure success, expecting operational answers like cycle time and error rate rather than innovation language. And ask what happens when a model changes, since anyone who has operated through a deprecation will say so.

Can one person do both roles?
Sometimes, for a narrow band of companies where marketing is the dominant function and the AI question is largely about visibility and how customers encounter the business. But the two are different skill sets, and an honest offer of both comes with explicit boundaries attached. An offering described without limits usually means the limits have not been found yet.

What should a fractional AI officer do in the first ninety days?
Map how work currently happens before recommending anything; identify which processes are genuinely suited to automation and which are not; set policy on data handling and on what must stay with a human; select tools against the mapped processes rather than in the abstract; train the people who will use them; and establish operational measures — cycle time, error rate, hours returned, adoption — so the programme can be judged on something other than enthusiasm.