Summary: Before hiring an AI consultant, ask 12 specific questions covering scope, pricing, experience, vendor independence, deliverables, and exit criteria. Good consultants answer in concrete terms, have references, quote fixed prices, don’t take kickbacks from AI vendors, and are willing to tell you when you don’t need them. If the person can’t answer most of these clearly, keep looking.
Why these questions matter
AI consulting is a new-enough category that anyone with a LinkedIn profile and a ChatGPT subscription can call themselves one. Most aren’t bad people — they’re just learning on your dollar. The questions below separate consultants with real practice from consultants who’ve read the same blog posts you have.
Ask them during the first discovery call. If the answers are vague, expensive, or hype-driven, that’s your signal.
The 12 questions
1. “What specific workflows have you built with small businesses?”
What you’re listening for: Concrete examples. “A 6-person real estate firm — we set up a listing-description generator and a follow-up email workflow that cut their response time in half.” Anonymized is fine; vague is not.
Red flag: Pure case-study language (“we unlocked 10x productivity across the enterprise”). Small-business consulting is a game of hours saved on specific tasks. Anyone who can’t describe specific tasks they’ve automated hasn’t done enough of the work.
2. “What will you not do in this engagement?”
What you’re listening for: Clear scope boundaries. “I won’t build custom software. I won’t touch your CRM configuration. I won’t train the model on your data.”
Red flag: “Whatever you need.” Unlimited scope is a sign the consultant hasn’t thought through delivery. You want someone with strong opinions about what’s in and out.
3. “Do you take any commission or referral payment from AI vendors?”
What you’re listening for: “No.” Short and declarative.
Red flag: Any long explanation. Some consultants are also resellers of specific platforms (Copilot, Salesforce Einstein, a specific automation tool). That’s not automatically bad, but it dramatically changes the advice you’ll get. You need to know.
4. “What’s included in the fixed price, and what would cause it to change?”
What you’re listening for: A specific deliverable list (“three written workflows, one team training session, one follow-up review call”) and explicit change triggers (“if you want to add a fourth workflow, it’s $X”).
Red flag: Hourly billing without a cap. Hourly at the small-business scale almost always goes over.
5. “How long will this take, and what’s the timeline week by week?”
What you’re listening for: A real schedule. “Week 1 — interviews and observation. Week 2 — draft roadmap. Week 3 — build the first workflow. Week 4 — iterate and hand off.”
Red flag: “Depends” or “we’ll see how it goes.” A consultant who’s done this before knows roughly how long each phase takes.
6. “What’s the smallest engagement you offer? What’s the smallest you’d recommend for my business?”
What you’re listening for: A real small-engagement option (a coaching session around $99) and a confident recommendation about whether yours is the right fit. Good consultants routinely tell prospects “start with the $249 session, not the $1,495 package.”
Red flag: Only expensive options. If the minimum engagement is $2,995, you’re talking to someone who’s priced out of your market — or who can’t do small work well.
7. “Can I talk to two or three past clients?”
What you’re listening for: “Yes, here are three — the first one is an architect in Durham, the second runs a property management firm, the third is a dental practice. Pick any of them and I’ll intro you.”
Red flag: Hesitation. Or “all my work is under NDA.” Occasional NDAs are real; a consultant whose every engagement is supposedly confidential usually doesn’t have recent engagements.
8. “How do you handle our data — and what happens to it when the engagement ends?”
What you’re listening for: Clear answers. “I use your existing tools; nothing leaves your environment. I don’t use your data to train anything. When we’re done, I delete my working notes. You keep everything.”
Red flag: Vague reassurance (“don’t worry about that”). Data handling is one of the most important things for a small business to get right, especially if you handle client information.
9. “What’s your technical background?”
What you’re listening for: Honest framing. “I’m not an ML researcher — I’m a former [role] who’s been using AI professionally for X years. I can configure and teach, not build foundation models.” Or similar calibration.
Red flag: Overselling. If every answer sounds like a pitch, this person is optimizing for impressing you, not helping you.
10. “Will my team actually be able to maintain what you build?”
What you’re listening for: Clear “yes, because…” answers. “Everything I build uses the tools you already pay for. The prompts and templates are plain text — anyone can edit them. I document everything so someone new could pick it up in an afternoon.”
Red flag: Dependencies that keep you chained to the consultant — custom platforms, proprietary scripts, undocumented configurations. The whole point of small-business AI is tools your team can own.
11. “What’s a recent engagement where you think you underdelivered?”
What you’re listening for: A real answer. Every experienced consultant has at least one. A good answer sounds like “I quoted a 3-week engagement that should have been 5 because I underestimated the client’s existing tool sprawl. I finished the scope, but I undercharged and I learned.”
Red flag: “I can’t think of one.” That’s either dishonesty or inexperience.
12. “What would tell you I don’t need a consultant at all?”
What you’re listening for: Someone willing to disqualify you. “If you haven’t spent at least a few hours using ChatGPT or Claude, you should do that first. If no one on your team is willing to spend an hour a week learning these tools, a consultant can’t fix that. If your problem is a messy process, AI will make it messier, not cleaner.”
Red flag: “Everyone can benefit from AI.” That’s a salesperson, not a consultant.
Bonus: three questions to ask yourself
- Does this person listen? If 80% of the first call is them talking, they’ll be the same way during delivery.
- Do I actually understand what they’re proposing? If you can’t explain it to someone on your team in 30 seconds, the proposal is too vague.
- Do they push back on anything I said? The best consultants disagree with you at least once on the first call. That’s a feature, not a bug.
Frequently asked questions
How long should the first call take?
Twenty to thirty minutes for a scoping call is standard. Anything under 15 minutes probably isn’t enough to be useful; anything over an hour is a selling tactic.
Should I expect a written proposal?
Yes — always. Even a $99 coaching session should come with a one-page description of what’s included, what it costs, and what you’ll walk away with.
Is it weird to ask about previous clients?
No. It’s the single most revealing question you can ask, and experienced consultants expect it. The awkwardness is on the consultant, not on you.
How many consultants should I talk to before deciding?
Two or three is usually enough. More than that and you’re usually procrastinating. The 12 questions above are a faster path than talking to 8 different people.
Is the free first call really free?
At the small-business tier, yes. Any consultant charging for a 20-minute scoping conversation is signaling low inbound demand. Paid discovery sessions make sense at the enterprise tier, not the SMB tier.
When this matters — and when it doesn’t
This applies if: You’re about to pay a consultant real money to help with AI, and you want to catch mismatches before the contract, not after.
Skip this if: You’re looking for free advice on Twitter or in a community. That’s a different conversation, and none of these questions really apply.
Related reading
Want to ask these questions on a real call?
I’m happy to answer every one of them in the first 20 minutes of a free scoping call. If the answers don’t match what you’re looking for, I’ll point you to someone who does.