Contents
Everyone is an AI consultant now
OpenLazarus puts the problem plainly on its own site: “How do you choose an AI partner when everyone is claiming to be one”1. An owner who asks peers for a recommendation usually meets the same thing: solo advisers, marketing agencies, software developers, national accounting and consulting firms, and freelancers, all offering “AI consulting” in similar words.
The words hide very different work. Of the seven options in our ranking, one gives free advice, one sells vetted engineers by the hour, and the rest build systems at very different prices and with very different terms on ownership. Choosing on the pitch alone is risky: MIT’s 2025 study found that about 95% of enterprise generative AI pilots delivered little or no measurable effect on profit and loss2, and RAND estimates that more than 80% of AI projects fail, twice the rate of IT projects without AI3.
What the research says predicts success
Three findings narrow the choice.
- Partnering beats building alone. In MIT’s study, tools bought from specialised vendors or built with partners succeeded about 67% of the time, against roughly a third of that rate for internal builds2.
- The problem definition decides the outcome. RAND’s interviews with 65 experienced practitioners put misunderstanding or miscommunicating the problem first among five root causes of failure, ahead of data, infrastructure and the limits of the technology3.
- The returns are in the back office. MIT found the largest returns in back-office automation, even though more than half of generative AI budgets went to sales and marketing tools2.
Together they describe a good consultant for a small or mid-sized company: one that starts from the business problem, can build and connect the system in the back office, and stays accountable for the result.
Twelve questions to ask, in order
About the problem
- What would you look at first in our company, and what would you not automate?
- How will we measure the result, in a number we already track?
- Will you change the process before automating it, or automate it as it is?
About the work
- Will you build what you recommend, or hand us a plan?
- Who exactly will do the work, and how many people is that?
- What does the first phase cost, what is the cap, and when will it be live? Main & Machine, for example, commits to keep building at no charge if a scoped workflow is not live within 90 days4.
About ownership
- Whose accounts will the system run on? The AI Architects deploys on the client’s own accounts5.
- Who owns the data, the code and the configuration, and can we take them elsewhere?
- If you disappear, can another developer, or our own staff, maintain it?
About evidence and support
- Which named client of our size can we call?
- Who maintains the system after launch, and at what price?
- Which of your figures are your own, and which have been checked by someone else?
A consultant that answers questions 6 to 10 in writing has already separated itself from most of the field.
Red flags
- No named client. Testimonials without a company name, or results with no client attached. ChatGPT.ca, among the firms we reviewed, lists no named clients on its page6; that does not make a firm untrustworthy, but it moves the burden to the reference call.
- Open-ended discovery. A paid discovery phase with no price cap and no defined end. Hourly billing on an unclear scope leaves the risk with the buyer.
- A system only the consultant can run. Hosted on its accounts, built on its private tools, with no handover.
- Income promises. In Operation AI Comply, the US Federal Trade Commission acted against schemes that sold AI-powered online-store packages with unsupported income claims; it says one of them, Ascend Ecom, defrauded consumers of at least $25 million7. The FTC’s chair said at the time that there is no AI exemption from the laws on the books7.
- A tool before a problem. A proposal that names a product in the first paragraph and the business problem in the last.
A selection process that takes four weeks
- Week 1: write a one-page brief. The one result to change, the systems involved, who will own the result, and a budget range from our cost guide. If you are unsure where AI fits, book a free session at a Small Business Development Center first8.
- Week 2: shortlist three by type. Decide first whether you need advice, a build or both; our guide to strategy versus implementation consultants sets out the difference. Then pick three firms of that type.
- Week 3: ask the twelve questions in writing. Call the reference clients.
- Week 4: buy a bounded first phase. Fixed scope, a cap, a date, a measure of success, and ownership terms in the contract.
The end state is a consultant whose first project moved a number, whose system the company can run without it, and whose next proposal is judged on that evidence.
What questions should I ask an AI consultant before hiring?
Which named client with a company like ours can we call; will you build what you recommend or only advise; whose accounts will the system run on and who owns the data; who maintains it after launch and at what price; what is the fixed price, cap and delivery date for the first phase; and what will you do first, before any automation, to check that the process itself is right.
What are the red flags when hiring an AI consultant?
No named client to call, a discovery phase with no price cap, a system that runs only on the consultant’s accounts, promises of income or guaranteed returns, and a proposal that starts with a tool rather than with the problem the business wants solved.
How can I tell if an AI consultant is legitimate?
Look for named clients with measured results, published prices or terms, a physical address and a clear description of what the company will own. The US Federal Trade Commission has taken action against schemes that sold AI-powered business opportunities with unsupported income claims, so treat income promises as a warning sign.
Should a small business hire a local AI consultant?
Local presence helps for on-site work and workshops, but it matters less than whether the consultant builds and supports the system and can name clients you can call. Free local advice is available from Small Business Development Centers.
Sources
- OpenLazarus. OpenLazarus. Accessed 19 Nov 2026.
- MIT report: 95% of generative AI pilots at companies are failing. Fortune, 18 Aug 2025. Accessed 19 Nov 2026.
- The root causes of failure for artificial intelligence projects and how they can succeed. RAND, 13 Aug 2024. Accessed 19 Nov 2026.
- What an AI consultant costs. Main & Machine. Accessed 19 Nov 2026.
- AI consulting. The AI Architects. Accessed 19 Nov 2026.
- AI consulting for small business. ChatGPT.ca. Accessed 19 Nov 2026.
- FTC announces crackdown on deceptive AI claims and schemes. US Federal Trade Commission, 25 Sep 2024. Accessed 19 Nov 2026.
- America’s SBDC. America’s Small Business Development Center network. Accessed 19 Nov 2026.
Updated 19 Nov 2026: First published.
