14 Questions to Ask Before Hiring an AI Consultant

Most AI consulting sales calls are a one-way pitch. The consultant walks you through their capabilities, you nod politely, and by the end you have a proposal but no real information. That is backwards - the hour is your best chance to filter, not theirs to close.

Here are the 14 questions that actually reveal whether a consultant is a good fit, grouped by what they test for.

Questions that test for domain fit

1. "What is a project you turned down recently and why?" Tests their judgment and whether they have enough volume to be selective.
2. "What kinds of problems are you consistently bad at?" Tests self-awareness. Everyone has weaknesses; a consultant who claims none is lying.
3. "What industries have you NOT worked in?" If yours is on the list, that may be fine - but you want to know.

Questions that test for technical honesty

4. "When does AI fail at what I am asking for?" Real consultants can name three failure modes immediately. Pretenders handwave.
5. "What would you build if the AI layer did not exist?" Sometimes the answer is "just use Zapier." A good consultant says so.
6. "Walk me through the actual prompts or logic you would use." Asks for specifics. If they can only talk in abstractions, they have not done it before.

Questions that test for business maturity

7. "What does success look like numerically?" Forces them to commit to measurable outcomes before you sign.
8. "How would I know in month 3 that this is going wrong?" Tests whether they think about monitoring and early warning, not just the happy path.
9. "What is the ongoing cost after launch - LLM calls, hosting, maintenance?" Many proposals lowball this. A real estimate has at least a monthly total.

Questions that test for engagement structure

10. "What happens if I want to stop working with you in month 4?" You should own the code, the documentation, the credentials. No lock-in.
11. "Do you offer fixed-price or only hourly?" For defined builds, fixed-price should be available. Hourly-only is a risk signal.
12. "Will one person own this, or will work be shuffled between team members?" Handoffs between consultants are where projects die. Know your point of contact.

Questions that test for references and real work

13. "Can I talk to a past client who did a similar project - and can you send me the live URL of something you built?" Logos on a website mean nothing. A URL and a conversation mean everything.
14. "Would you work on this project for a share of the outcome instead of a fee?" Nobody will say yes - but the hesitation and how they explain their "no" tells you everything about their confidence in the outcome.

The best consultants I know get asked fewer of these questions than average - because the clients who have learned to ask them usually already know they are a good fit.

What to do with the answers

There is no perfect score. The signal is pattern-level: a consultant who handles most of these 14 questions with specifics is almost always a safer hire than one who handles most of them with marketing language. You are looking for "I thought about this already" energy, not "impressive pitch" energy.

One final filter

At the end of the call, notice how you feel. Relieved? Confused? Sold-to? Tired? That emotional read is data. A good consultant leaves you with more clarity than you started the call with, not less.