How to Choose an AI Consultant: 12 Filters That Separate Signal From Noise

A friend who runs a $40M services company called me last quarter: "I just paid $90,000 to an AI consultant for a strategy deck. The deck is beautiful. My team has no idea what to do with it."

That sentence is the AI consulting market in one line. A lot of consultants are selling decks. Very few are selling transformation. The difference shows up in months 3-6, when the deck is in a drawer and the team is asking the same questions you started with.

The twelve filters below are how you separate one from the other. They work whether you're hiring a solo advisor or evaluating a Big 4 practice. They are not subjective. Each one has a specific answer the consultant either has or doesn't have.

Filter 1 — Do they have a named framework?

Real consultants have a method with a name, a structure, and case studies. "We tailor to each client" is consultant-code for "we don't have a method, we'll invoice you while we make one up." The consultant whose method is named, written, and applied across multiple clients has built something. The one who can't name their method is selling time.

What to ask: "What's your method called? Can I see the framework on one page?" If the answer takes three minutes to deliver, you have the right consultant. If it requires a 40-slide deck to explain, walk.

Filter 2 — Have they shipped, not just advised?

The AI consulting market has a heavy population of advisors who have never operationalized anything. They've read the papers, attended the conferences, learned the vocabulary. They can talk about AI elegantly. But they've never had to deal with the actual mess of an enterprise AI rollout: legal review, data perimeter, team change-management, vendor disputes, model drift.

What to ask: "Name three companies where your work moved from strategy deck to working operation. What changed in their metrics?" Vague answers = no shipping experience. Specific answers with numbers = real operator.

Filter 3 — Do they refuse projects?

The consultant who says yes to everything is selling availability, not judgment. The one who's worth hiring will refuse projects where the readiness isn't there — and tell you what pre-work is required first. This is counter-intuitive: a consultant who turns away revenue is signaling integrity, not weakness.

Test: Describe an early-stage AI use case that's clearly not ready (no data, no team capacity, no problem clarity). If they say "we can make it work," they'll burn your budget. If they say "you need to do X, Y, Z first before this is a real project," they're worth hiring.

Filter 4 — Can they explain it to your board in plain language?

If the consultant slips into "vector embeddings," "RAG architecture," and "fine-tuning thresholds" during the first 30 minutes with you, they'll be worse with your board. The discipline of explaining complex systems in plain executive language is a separator. The consultant who has it builds organizational alignment. The one who doesn't builds a knowledge silo around themselves.

What to ask: "How would you explain our project's risk to my board in two minutes?" The good consultant explains it. The bad one says "well, it's complex."

Filter 5 — Do they bring written governance, or just enthusiasm?

The Brown Belt of AI Black Belt is explicit on this: every serious AI implementation requires governance — named accountability, written perimeters, audit cadence, escalation triggers. The consultant who delivers a strategy without governance design is delivering half the work. You'll feel the missing half at month four when the system makes a bad decision and no one knows who's accountable.

What to look for in their proposal: explicit sections on accountability matrix, decision perimeters, audit framework, kill-switch protocol. If those are absent, ask why. The deeper governance piece, for context, is AI Governance Leadership.

Filter 6 — Are they vendor-neutral?

Many consultants have undisclosed referral relationships with specific AI vendors. They get paid by you AND by the vendor when they recommend a product. This is a structural conflict of interest. The vendor-neutral consultant evaluates options against your needs, not against their commission stack.

What to ask: "Do you have referral, partnership, or commission relationships with any AI vendors? Disclose all of them in writing." The honest consultant discloses. The dishonest one says "no, we're fully independent" and then somehow always recommends the same three vendors.

Filter 7 — Do they understand your industry's constraints?

Healthcare AI has HIPAA, financial-services AI has SOC 2 and regulatory disclosure, legal AI has privilege concerns, manufacturing AI has IP and operational-technology overlap. The consultant who doesn't know your industry's regulatory and operational constraints will design solutions that get rejected at month two by your legal or compliance team.

Test: Describe one regulatory or operational constraint in your industry. Watch their response. If they handle it fluently, they understand your domain. If they say "good question, we'll dig into that," they're learning on your dime.

Filter 8 — What's their AI failure rate, and what did they learn?

Consultants who claim 100% success rate are either lying or new. Real operators have failures and have learned from them. Asking for their failure stories is the fastest way to surface the consultants worth hiring — they answer crisply, with specifics, and tell you what they changed.

What to ask: "Tell me about an AI project you led that didn't work. What did you change in your method afterward?" The good consultant has a clear answer. The bad one deflects or claims they've never failed.

Filter 9 — Do their contract terms protect your data and your exit?

The contract is where consultants reveal who they really are. Look for: clear data ownership (yours, not theirs), clear deliverables (in writing, not "we'll align"), clear exit terms (you can leave at month 3 without penalty if it's not working), clear IP terms (any models or frameworks built on your data are yours).

Red flags: auto-renewal clauses, vague deliverables, expansive non-compete language, ambiguous IP ownership, "right to use your case as a case study" without your explicit approval per case.

Filter 10 — Do they have a presence beyond LinkedIn?

LinkedIn-only consultants are everywhere. The ones worth hiring have a substantive presence: a book, a method documented in long-form writing, conference keynotes with the slides public, a podcast or newsletter where their thinking is on display. This isn't vanity — it's evidence that their method has been pressure-tested in public, not just in slide decks.

The deeper piece on what executive AI thought leadership actually looks like is Hapkido Business Principles.

Filter 11 — Do they bring a team or just themselves?

Solo consultants can be excellent for advisory engagements. For implementation, you need a team — typically a senior advisor (the consultant you met), an operational lead (who runs the day-to-day), and a technical liaison (who interfaces with your IT). The consultant who says "it's just me" on an implementation engagement is either understaffing the project or planning to bill you for sub-contractors at full rates.

What to ask: "Who specifically will be on this project? What % of their time? What's their hourly rate vs. yours?" Get names, percentages, and rates before signing.

Filter 12 — Do they invest in your team's learning, or hoard the knowledge?

The bad consultant creates dependency: they're the only one who understands the system, so you can never leave. The good consultant trains your team to operate independently — and is happy to do so, because their reputation is built on the result, not on the lock-in.

Test: Ask explicitly how the engagement transfers knowledge to your internal team. The good answer has structure: documented playbooks, paired sessions, an internal champion designated, a measurable hand-off milestone. The bad answer is "we'll always be here when you need us."

Frequently asked questions

When you're ready

If you're evaluating AI consultants and want to compare them against the twelve filters in a discovery conversation, executive AI consulting is structured exactly this way — framework first, governance second, vendor-neutral throughout, with explicit knowledge transfer to your team.

For executive boards or leadership offsites, the keynote speaking format delivers the twelve filters as a 60-minute working session your team can apply to their actual vendor shortlist.

For the structural piece on what makes corporate AI initiatives succeed or fail at the system level, read Corporate AI Implementation Failure. For the executive-decision checklist that precedes the consultant question, read AI Implementation Checklist for Executives.

Want the complete method — the seven belts, the named frameworks (AMARTE, Hwa·Won·Ryu, Tumanov Filter, Green Matrix, PAF, PMP Triangle, Master Map of AI Systematization), and integrated case studies? Read AI Black Belt: Fundamentals Before the Prompt. Available now on Amazon in Spanish; English edition in final author review.