How to Hire an AI Consultant — and the 5 Red Flags

Eighteen months ago, "AI consultant" wasn't a job title. Now it's a gold rush. The problem with gold rushes is that most of the people selling shovels have never dug. The market is flooded with consultants who can talk fluently about AI — the trends, the models, the future of work — and have never actually shipped a working system inside a real business.

That's the core thing to screen for, and it's the through-line of everything below: the gap between people who talk about AI and people who build with it. Talkers are abundant and confident. Builders are rarer, quieter, and have receipts. Here's how to tell which one is sitting across from you, and the five red flags that give the talkers away.

What you're actually hiring for

Be clear on the job before you evaluate candidates. You are not hiring someone to explain AI to you — you can get that from a podcast for free. You're hiring someone to change how your business operates: to take a process that costs you time or money and rebuild it so AI does the heavy lifting, then prove the result in numbers you care about. That's an execution job, not a thought-leadership job. Evaluate accordingly.

The best AI consultants are operators first. They've run something — an agency, a team, a product — and learned to use AI because they had real work that needed leverage. Their knowledge is downstream of doing. The weakest are commentators first: they learned to talk about AI as the product itself, and the "consulting" is just the talk, monetized.

The 5 red flags

Any one of these should slow you down. Two or more, walk away.

1. They can't show you something they've shipped. Ask, "What's a system you built that's running right now, and what did it change?" A builder answers with a specific story — the process, the before, the after, the number. A talker pivots to strategy, frameworks, and what's "possible." If they can't point to a live thing they made, they make slide decks, not systems.
2. Everything is "AI-powered" and nothing is specific. Vague, buzzword-dense language is the tell. "We leverage cutting-edge AI to transform your business" means nothing. A real one talks in concretes: this workflow, this tool, this integration, this many hours saved. Specificity is the signature of someone who's actually done the work; abstraction is the camouflage of someone who hasn't.
3. They sell the model, not the outcome. If the pitch is about which AI is the smartest and the demo is a chatbot answering trivia, they're selling technology theater. The model is a commodity. The outcome — a process that now runs itself, a cost that dropped, a thing that ships faster — is the product. Beware anyone more excited about the tool than about your result.
4. No plan to transfer it to you. Ask how your team will run what they build. If the answer keeps the system locked in their tools and their heads, you're buying a dependency, not a capability. Good operators document and hand off because they're confident the work stands on its own. (This is the same lock-in trap that plagues marketing agencies — same tell, different industry.)
5. They promise transformation in two weeks and no honest caveats. Anyone who tells you AI will revolutionize your whole business overnight, with no mention of the upfront work, the messy data, or the things that won't automate cleanly, is selling a fantasy. Real implementation is a sequence of unglamorous wins. The honesty about what's hard is itself a credibility signal.

The questions that separate builders from talkers

Bring these to the first call. The quality of the answers will sort the field fast:

• "Walk me through a system you've built end to end." Listen for specifics and friction — real builds have war stories. Smooth, abstract answers are a flag.
• "What's something AI couldn't do well that you expected it to?" Builders have scars and will name them happily. Talkers have no failures because they've never shipped anything to fail.
• "If I hired you, what's the first workflow you'd touch and why?" A good answer is grounded in your business within minutes. A bad one stays generic because they think in templates, not operations.
• "What do I keep if we stop working together?" The asset-vs-dependency question. The reaction matters as much as the words.
• "Show me your own operation." The most honest tell of all. People who run their own business on AI can show you. People who only consult about it can't.

Green flags worth paying for

It's not all warnings. Here's what a strong AI consultant looks like, so you know what you're aiming for:

• They run their own business on the same systems they'll sell you. The most credible proof there is. They've eaten their own cooking, and it's why they trust the recipe.
• They talk about your problem more than about AI. The technology is a means; your outcome is the point. They get curious about your operation fast.
• They're specific about limits. They'll tell you what won't work, what's not worth automating yet, and where a human still beats the machine. Confidence with honest boundaries beats hype every time.
• They sequence the work. One workflow at a time, each win funding the next, instead of a big-bang "digital transformation." That's how real implementation actually goes.
• They leave you more capable. Their goal is for your business to own the capability, not to keep you renting it.

The one-sentence test

If you only remember one thing: hire the operator who can show you a system they built and run, not the commentator who can describe one. The market is loud with people who talk about the future of work; the value is with the small number quietly building it. Make them show you the build. The ones with something real to show will be glad you asked — and the ones without will tell you everything you need to know by how fast they change the subject.