10 Questions to Ask Before Hiring an AI Consulting Firm

Most AI consulting sales calls go one way: the vendor talks, the prospect nods, the demo lands, and the conversation never gets to the questions that actually matter. This is the list we wish every prospect asked us — the questions that separate a competent partner from a vendor selling a capability they don't understand well enough to support.

The ten questions

1. What have you built for a business like ours? Vague answers or wildly different verticals are a flag.
2. Who does the build? Is it the same people I meet in the sales call?Watch for firms that sell you a senior partner and hand the work to a junior team offshore.
3. Are you fixed-scope, fixed-price or time-and-materials?Either can work, but you should know before you sign. See What "fixed-scope, fixed-price AI" actually means.
4. What happens if the engagement runs long? A good answer is specific. A vague answer is expensive.
5. What's your regulatory posture? HIPAA, IRS 7216, attorney-client privilege, CUI, fair housing — the answer should be specific to your situation.
6. Do you train models on client data? The answer should be no and it should be contractual. See HIPAA BAAs for AI vendors.
7. Where does the model run? Public inference, dedicated tenant, private cloud, on-premise — the right answer depends on your data.
8. What does week-by-week look like? A good vendor can tell you what happens in week 1, week 3, week 5 of your engagement.
9. Who tunes the system after cutover? AI deployments need tuning. If the vendor disappears after launch, performance plateaus.
10. Can you share references in our vertical? Any competent firm should be able to produce at least one.

The answers we give

For transparency, our answers: we work mostly with DMV SMBs across healthcare, legal, tax, real estate, home services, and professional services. The work is done by us — not offshore, not a subcontracted team. We are fixed-scope, fixed-price. If the scope changes mid-engagement, we re-scope and re-quote before doing the work. Regulatory posture is specific to each engagement. We do not train models on client data and it's in the contract. Model deployment varies by data sensitivity. The engagement arc is typically 2-6 weeks, tuning continues in the ongoing-partner model, and we'll connect you to references in your vertical on request.

What vague answers look like

• "We're HIPAA-compliant" without specifying architecture.
• "We use best-in-class AI" without naming the model or deployment.
• "We'll figure out scope after we start" without a written SOW.
• "We can't share references" with no clear reason.

Ask us any of these questions on a scoping call — we'll answer each one specifically.