How to choose an AI consultant — seven questions that separate builders from deck-makers

The market for AI help is loud, and most of it sounds identical. Everyone promises transformation, everyone has a framework, and everyone's case studies are anonymised. The hard part isn't finding someone who'll take the work — it's telling apart the people who will build you something that runs from the people who will sell you a strategy and disappear before anything reaches production.

Here are the questions that draw that line. Ask them on the first call.

1. "Will you build it, or advise us on building it?"

This is the fastest filter. Some firms deliver a solution running on your systems; others deliver a recommendation and leave the implementation risk with you. Neither is wrong — but you need to know which you're buying, because a strategy deck priced like a delivery engagement is the single most common way AI budgets get wasted.

2. "What KPI would we agree to move, and by how much?"

A serious partner will want to anchor the work to a number your business already tracks before they quote. If the answer stays at the level of "improve efficiency" or "unlock insights," they haven't scoped a project — they've scoped a phase-one discovery that will scope the real project later, on your dime.

3. "Whose cloud does this run on, and where does our data live?"

The right answer, almost always, is yours. If a consultant's approach requires your data to leave your environment or land in a platform you don't control, that's not a technical detail — it's a compliance, cost, and lock-in decision you're making by accident. Make it on purpose.

4. "How do you keep a human in the loop?"

In healthcare, legal, and security, the value isn't full automation — it's taking the slow first pass off an expert's plate while they stay in control of every decision. A consultant who talks about removing people rather than accelerating them either doesn't understand your risk surface or is selling you a headcount argument you didn't ask for.

5. "Can we start small and stop?"

Beware the single large engagement that only pays off at the end. The lower-risk structure is stages: each one stands alone, ends in something real, and is yours to continue based on whether the number moved. If a consultant can only sell you the whole thing at once, they're asking you to carry all the risk up front.

6. "What happens if the feasibility check says no?"

Ask directly. A consultant worth hiring will tell you they'll say "not yet, and here's the one thing that has to change first" — and mean it. One who treats every problem as an AI problem will find a way to bill you regardless of whether AI was ever the right tool.

7. "Who, specifically, will do the work?"

Find out whether the people on the sales call are the people who'll write the code. In small, hands-on teams they're the same. In larger shops the gap between the partner who pitches and the junior who delivers is where quality quietly leaks out.

The best AI consultant makes themselves easy to stop and easy to verify. Anyone who needs a long contract and a leap of faith is managing their risk by transferring it to you.

The short version

You're not buying intelligence — you can rent that anywhere now. You're buying accountability: someone who will name the number, build the thing, keep your data where it belongs, and tell you the truth when the answer is no.

If those are the answers you're looking for, book a discovery call. We'll run your slowest process through exactly these questions — and if AI isn't the right fit yet, we'll tell you that too.