7 Questions to Ask Before You Hire an AI Consultant

There are two types of people calling themselves AI consultants right now. The first has spent years deploying production AI systems, has a clear methodology, can articulate exactly what they'll build and why it will work, and has clients who'll tell you what changed after the engagement. The second has been reading about AI since ChatGPT launched, built some impressive demos, and is confident enough to charge consulting rates.

Both types give the same kind of presentation. The questions below separate them.

Question 1: "Can you walk me through a specific deployment that failed, and what you learned from it?"

Good answer: a real story, with specifics — what they built, why it didn't work as expected, what they changed, and what the outcome was eventually. Anyone who has deployed production AI systems has failure stories.

Red flag: they haven't encountered significant failures yet, which means they haven't done enough deployments. Or they tell you about client failures that were the client's fault. Neither is a satisfying answer.

Question 2: "What does your data assessment process look like before you start any work?"

Good answer: a structured evaluation of data quality, availability, format, and completeness before scope is agreed. Real AI projects fail on data quality more often than on algorithm choice. A consultant who doesn't assess this first is working blind.

Red flag: "We can work with what you have" without explaining what assessment will tell you about what you actually have.

Question 3: "What percentage of your proposals result in you recommending not to proceed?"

Good answer: some meaningful percentage — 10%, 20%, more. A consultant who has never told a client the project isn't viable yet hasn't had enough honest conversations about whether AI is the right solution.

Red flag: "We always find a way to make it work." That's a salesperson's answer.

Question 4: "How do you handle the handover after the project? Who maintains it?"

Good answer: a specific plan — either ongoing retainer, internal documentation sufficient for your team to maintain it, or a named team member trained to manage it. This is a question most clients don't ask and most consultants haven't thought through carefully.

Red flag: vague reassurances about "comprehensive documentation" with no details about what that means in practice.

Question 5: "What tools do you typically recommend, and what are their weaknesses?"

Good answer: named tools with genuine trade-off analysis. No tool is right for every situation, and a consultant who knows their field will tell you what each option can't do as readily as what it can.

Red flag: enthusiasm without caveats. Every tool has weaknesses; someone who can't name them doesn't know the tool well enough.

Question 6: "What does success look like for this engagement, and how do we measure it?"

Good answer: specific metrics agreed before work starts — not general outcomes like "improved efficiency" but defined numbers: resolution rate, processing time, error rate, revenue impact. Without these, there's no way to evaluate whether the engagement worked.

Red flag: "That depends on what you want to achieve" without following up to actually define it.

Question 7: "Can I speak to two clients from projects similar to mine?"

Good answer: yes, followed by actual introductions. Not "I'll ask them if they're comfortable" as a delaying tactic, but a genuine offer to connect you with references who'll speak candidly.

Red flag: any hesitation. If a consultant has genuinely delivered strong results, their clients want to tell people. Reluctance to provide references is information.