How to Vet an AI Advisory Firm Before You Sign

Choose an AI advisor by matching the type of provider to the type of problem, then vetting for production track record, senior attention, and a plan for what happens after they leave. The stakes of choosing well are rising: over the trailing 90 days, versus the same window a year earlier, demand for AI and machine learning advisory grew 167% as a share of all projects private equity firms brought to BluWave, per BluWave project data. Supply has responded with a flood of newly minted 'AI advisors' of uneven quality.

One thing to know before reading any guide on this subject, including this one: nearly every "how to choose an AI consultant" article on the internet is written by an AI consulting firm. BluWave does not sell AI consulting; it matches companies to independently vetted providers under a published no-pay-to-play policy, meaning providers cannot pay to be included or ranked. What follows is the evaluation logic that vetting work produces.

What do AI advisors actually do for mid-sized companies?

An AI advisor's job is to turn a vague mandate to do something with AI into a short list of use cases with owners, costs, and payback periods, and then to get the first one working in production. The category spans four different animals: strategy firms that build the roadmap, implementation shops that build the systems, data readiness specialists who fix the data plumbing first, and fractional AI executives who own the agenda inside the company a few days a week.

Matching the type to the problem is most of the decision, and each type gets vetted differently.

• If you don't yet know what you want: buy assessment and roadmap work from a strategy advisor, and weigh references at your company size most heavily. Enterprise frameworks do not shrink well.
• If you know the use case but the data is a mess: start with a data readiness specialist, and vet the unglamorous specifics, such as pipelines, permissions, and cleanup estimates that survive contact with your actual systems.
• If direction is set and you need it built: use an implementation shop, and vet production track record above all.
• If nobody inside the company owns the agenda: a fractional AI executive comes first, and that vetting is a hiring decision, with references from the operating seat, not the conference stage.

Buying implementation before direction is how mid-sized companies end up with tools nobody uses.

The eight questions to ask before hiring an AI advisor

Ask every candidate the same eight questions and compare the answers side by side.

1. What have you put into production for a company our size, and is it still running? Demos and pilots do not count. The follow-up that separates operators from salespeople: what broke after launch, and who fixed it?
2. Who exactly will do the work? The senior person who pitches is often not the person who delivers. Ask for the names, their backgrounds, and how much of their time you get.
3. What does the readiness assessment cover, and what does it cost if we stop there? A credible advisor will assess data, systems, and people before recommending anything, and will price the assessment so you can stop there.
4. How will you prioritize use cases? The answer should involve payback periods and process owners. If every answer is "agents," keep interviewing.
5. What will this cost us beyond your fees? Data cleanup, software, and your team's time are usually the larger investment. An advisor who has done this before will estimate all three unprompted.
6. Do you take referral fees from, or resell, any platform you might recommend? Many advisors are also resellers. That is not disqualifying, but it must be disclosed, and a "platform-agnostic" claim should survive the follow-up: describe a recent engagement where you recommended against the tool you resell.
7. Who owns what you build, and what do we depend on you for afterward? You should own the work product outright, and the engagement should build your team's capability rather than a permanent dependency.
8. What would make you tell us not to do this? The best advisors can describe the client they turned away. If there is no such story, the advice is a sales channel.

Should we hire someone for AI or bring in an outside firm?

Bring in outside help to set direction; hire when there is a full-time job to own. Below a certain scale, a dedicated AI executive is hard to justify. The common sequence in the lower middle market is an advisory engagement to build the roadmap, then a fractional AI leader to execute it a few days a week, then a full-time hire only when usage and pipeline justify one.

The wrong answer is usually hiring first. A new AI leader with no roadmap inherits the "figure out AI" problem with a salary attached, and the role fails for the same reason unscoped advisory engagements fail: no defined use cases, no owners, no payback clock.

What does AI advisory cost?

Cost depends on scope and follows the sequence: a focused readiness assessment sits at the low end, strategy-through-pilot in the middle, and buildout with systems integration at the top. Whatever the number, weigh it against the cost of the alternative, which is buying tools before defining use cases and writing off both the spend and a year of organizational patience.

Two cost questions matter more than the headline fee: what the total investment is once data cleanup, software, and internal time are counted, and whether you can buy the readiness assessment alone before committing to more. An advisor who resists the second question is telling you something.

"The providers who survive our vetting share a tell," says Keenan Kolinsky, Head of Research & Operations at BluWave. "They can name the client they told not to buy. Selection pressure works the same way for you: the advisor willing to shrink the engagement is the one you can trust with a bigger one."

Frequently asked questions

Our board keeps asking about our AI strategy. Where do we start?

Start with a readiness assessment from an advisor who will sell it to you as a standalone piece of work. It gives you a defensible board answer grounded in your own data, systems, and team, and it tells you whether the bigger engagement is worth buying at all.

How do we pursue AI without overextending the budget or the team?

Scope narrow and deep: one or two use cases where the process is understood, the data exists, and a manager owns the outcome. The budget failures in mid-sized companies come from taking on too many use cases at once.

How long does an AI advisory engagement take?

Readiness assessments and proof-of-concept work often run 6-12 weeks, and strategy roadmap engagements 8-16 weeks, per BluWave's published guidance. Implementation is scoped by use case, from a few months for a focused automation to much longer for portfolio-wide work.

What are the red flags when evaluating an AI advisor?

The biggest red flags are no production references at your company size and a pitch that starts with technology instead of your processes. Add resistance to selling a readiness assessment as standalone work, vagueness about who will actually do the work, and no story about a client they advised not to proceed.

How do we know if a provider has actually done this before?

References from companies your size, in production, still running. BluWave's approach to this is structured vetting: application review, reference checks with prior PE clients, and re-vetting for each specific engagement rather than a one-time stamp.

The next step

If you would rather start from a vetted shortlist than from an open market of self-described AI experts, that is the problem BluWave exists to solve: describe your need, and get matched to AI Tools & Services BluWave Vetted™ providers whose production track record has already been checked, within 24 hours and at no cost to connect.