AI consultant or fractional CTO: which one does the problem need?

Both titles turn up in the same search, and both people will take the call. They’re hired to answer different questions, and the expensive mistake is hiring one to answer the other’s.

What each one is hired for

An AI consultant is hired for a technology. The brief starts from the tool: we should be using AI, tell us where. That’s a fair brief when the business already knows its own problems and wants someone who knows the tooling better than anybody on the payroll.

A fractional CTO is hired for a business. The brief starts from the company: the technology is holding us back, or we can’t judge what we’re being told, and someone senior needs to own it. AI is one of the things that person may reach for. So is a rewrite, a hire, a supplier change, or doing nothing for six months.

The difference shows in the first recommendation. Ask an AI consultant where AI fits and you’ll get a list of places AI fits. The question has already assumed its answer.

The test

Ask yourself what you’d do if the answer came back as “no AI at all”.

If that would be a useful result, because it frees the budget for something else, you have a business problem and you want the person who owns business problems. If it would feel like a wasted engagement, you have already decided to build with AI and what you need is a good builder.

Both are legitimate. The first is more common than people admit. A fair share of what reaches me as an AI project turns out to be a search box, a report nobody wrote, or an approval chain with too many people in it. None of those needs a model, and all of them are cheaper to fix than the project that was proposed.

When an AI specialist is the right hire

Hire the specialist when three things are true.

You know which workflow you want changed, and you can describe it. You know what a wrong answer costs and who would notice. And you have somebody inside the business who owns the technical direction, so the specialist’s work lands in an architecture somebody is looking after.

With those in place, a specialist is faster and you shouldn’t pay for breadth you won’t use.

When it’s the wrong hire

The third condition is the one that goes missing. A business with nobody owning the technology hires an AI consultant, gets a working pilot, and then finds there’s nobody to run it, nobody to say how it fits with the systems around it, and nobody to decide what happens when the model changes underneath it.

That pilot was never an AI problem. It was a leadership gap with an AI project dropped into it.

The opposite mistake exists too. A fractional CTO who can’t build will hand the AI work to somebody else and manage it from a distance, and you’re now paying two people to have one conversation. If the AI work is a serious part of the plan, ask whoever you’re talking to what they have in production, what it costs per month to run, and how they know when it’s wrong. Those three answers separate people who have shipped it from people who have read about it.

Where I sit

I do both, and I’d be wary of that sentence from anyone else, so here is what sits behind it.

I’m the CEO of a business I co-founded, and I lead its technology. The AI in that business runs on hardware I own because it handles children’s data, the local model was chosen by running it against a cloud model over identical input, and every model call on the platforms I build is logged against a budget. On a client platform, the AI’s answers about money are checked against the ledger to a tolerance of one penny.

I still start from the business. AI is a tool on this site and not the category, and the first thing I want to know is what the work is supposed to change commercially.

What to do with this

If you can’t yet say which workflow should change, don’t hire anybody to build. Find out what is costing you first. That’s a two-week job, and it’s the one I’d start with whatever title is on the door of whoever you hire.