Buyer guide
How to choose an AI consultant
The whole decision comes down to one thing: can they show evidence of work that shipped, or only talk about it? This is the short version of how we vet the roster, turned into questions you can ask on the call, red flags that should end it, and a checklist you can run yourself.
The same test we apply, in your hands.
On the call
What questions should I ask an AI consultant?
Seven questions, in rough order. The answers matter less than whether they are specific and unrehearsed.
- 1
Can you walk me through a project like mine, start to finish?
A real implementer tells a specific story: the problem, what they built, the guardrails, and the measured result. A reseller talks in generalities and pivots to their process deck.
- 2
Who owns and maintains the system after go-live?
Most AI projects fail three months in, when nobody owns the automation and it drifts. Get the answer, and the cost of it, before you sign.
- 3
What happens when the system is unsure or gets it wrong?
Good work escalates to a human, logs the failure, and fails loudly. If they cannot describe the failure mode, they have not run this in production.
- 4
How do you measure that it actually worked?
Ask for the metric before the demo. A build tied to resolution rate, hours saved, or conversion is accountable. A build sold on a flashy demo is not.
- 5
Show me a named client and a number I can check.
Vague promises of results are free. A named engagement with a before-and-after figure is evidence. If they cannot produce one, ask why.
- 6
How do you handle my data, access, and anything sensitive?
In any regulated or data-heavy context, a serious partner raises this before you do. Silence here is disqualifying.
- 7
Is this a fixed scope, and what does change cost?
A written scope with a clear price protects you from open-ended billing. Watch for anyone who wants hourly on a large build with no cap.
Walk away if
What are the red flags when hiring an AI consultant?
Case studies that name no client, no metric, and nothing you can verify.
A certification that turns out to be a badge bought from a directory, not a real vendor partnership.
A flashy autonomous demo with no mention of guardrails, evaluation, or how they measured it works.
No answer for who maintains the system after launch, or what that costs.
Pricing quoted sight-unseen, or open-ended hourly on a big build with no cap.
All audience and thought-leadership, no shipped work for a paying client.
Reluctance to put scope, ownership, or data handling in writing.
Before you sign
What evidence should I check before hiring?
The five independent signals we vet the roster on, framed as things to check yourself. Two or more is a genuinely strong candidate.
- 1A real vendor partnershipA genuine partner listing (Claude, Zapier, Make, n8n, a cloud provider), not a badge bought from a certifying directory.
- 2A named case study with a metricA specific client and a concrete before-and-after number. Self-hosted is fine, so long as it is checkable.
- 3Third-party reviews you can readFive or more on Clutch, G2, Google, or Upwork, with the rating visible.
- 4Cross-listings in other directoriesCorroboration across two or more independent directories. Rare, and meaningful when it is there.
- 5A public trail of real workOpen source, published templates or workflows, talks, an audience built on demonstrated work.
This is exactly how our own vetting works. Read the full methodology.
Common questions
Choosing an AI consultant: common questions
What questions should I ask an AI consultant?
Ask them to walk through a project like yours end to end, who owns the system after launch, what happens when it is unsure or wrong, how they measure success, and to show a named client with a checkable number. Then ask about data handling and whether the scope and price are fixed. The specific, unrehearsed answers separate operators from resellers.
How do I know if an AI consultant is legit?
Look for evidence you can check rather than claims you cannot. Two or more of these is a strong signal: a real vendor partnership, a named case study with a metric, third-party reviews, cross-listings in independent directories, and a public trail of shipped work. A polished website with none of that behind it is the classic label-slapper.
What are the biggest red flags when hiring an AI consultant?
Unverifiable case studies, purchased certification badges, flashy autonomous demos with no guardrails, no plan for who maintains the system, pricing quoted before scoping, and all audience with no shipped client work. Any one of those is a reason to slow down; several together is a reason to walk away.
Do I need an AI consultant, or can I do it in-house?
It depends on whether you have the capacity and the specialized skill in-house, and how fast you need it. A consultant buys speed and focused expertise without a permanent hire; an in-house team builds lasting capability and owns the result. The consultant-versus-in-house comparison lays out the trade-offs side by side.
How many AI consultants should I talk to before deciding?
Enough to compare, usually two or three, but judged on evidence rather than pitch quality. The point is not to collect proposals; it is to find the one who can show shipped, maintained work that resembles what you need, and who answers the hard questions without deflecting.
Consultant, agency, or in-house?
If you are still deciding whether to hire out at all, the trade-offs are laid out side by side.
Think a firm belongs on the roster and can point to the evidence? Suggest a listing.