How to Choose an AI Consultant (12 Questions to Ask Before You Sign)

1. "What will be running when you leave?"

Good answer: a specific list. "An intake automation that creates CRM records from your web form, an internal assistant answering from your SOPs and a trained team of 25." Worry: "A strategy," "a roadmap" or "recommendations" as the only deliverable. A roadmap is useful as step one. It shouldn't be the whole engagement.

2. "Can you show me something you built that's running today?"

Good answer: a live demo, a screen recording or a walkthrough of a real automation or assistant, with client details removed. Worry: only slides, mockups or "we can't show client work." Confidentiality is real, but a consultant who builds things can find something to show.

3. "Which AI tools do you recommend, and do you earn anything from them?"

Good answer: "It depends on your stack," followed by sensible reasoning, and a clear statement about commissions or reseller margins. Worry: the same tool for every client, or evasion on commissions. Reseller arrangements aren't automatically bad, but you should know about them.

4. "Who exactly will do the work?"

Good answer: named people and their roles, plus how subcontractors or partners are used and who stays accountable. Worry: a senior person who sells the deal and disappears, or unnamed offshore teams with no accountable lead.

5. "Whose accounts will everything be built in?"

Good answer: "Yours. You'll have admin access to every tool and login." Worry: automations in the consultant's accounts, or a proprietary platform only they can maintain. That's lock-in.

6. "What happens to our data?"

Good answer: which tools will see what data, whether those tools train on it, how long it's kept and what agreements are needed. Business plans from the major vendors say they don't train on business data by default, but details vary. Worry: "It's all secure," with nothing specific.

7. "How will we measure whether this worked?"

Good answer: a baseline measured before the work starts, such as hours per task or tasks per week, and a report against it afterward. Worry: promised percentages before anyone has looked at your work. "Clients see 40% productivity gains" is a sales line, not a measurement.

8. "What does it cost, all in?"

Good answer: a fixed project price or a clear retainer, plus a separate estimate for licenses and monthly running costs. Worry: hourly billing with no cap on a defined project, or licenses bundled into a single number. Our AI consulting cost guide covers typical ranges.

9. "What's your experience with businesses our size?"

Good answer: examples of work with companies of similar size and budget, even if anonymized. Worry: only enterprise examples. A firm used to 18-month programs for corporations often doesn't fit a 40-person company that needs results this quarter.

10. "What would you tell us not to do?"

Good answer: specifics. "Don't buy Copilot for your field crew." "Don't automate customer replies without a review step yet." "Fix your file permissions first." Worry: nothing. A consultant who never says no is selling, not advising.

11. "How do you train our people?"

Good answer: training built on your team's real tasks, by role, with follow-up after a few weeks. Worry: a one-hour webinar, or no training at all. Tools that people don't use save nothing.

12. "What happens if we stop working together?"

Good answer: documentation, a handoff session and full ownership of everything built. Worry: hesitation, fees to export your own work or a platform that stops working when the contract ends.

Give each answer a score:

A strong candidate scores 20 or more out of 24. Below 16, keep looking. Treat any zero on questions 2, 5 or 6 as disqualifying. Those are the ones that cost you most when they go wrong.

Ask each candidate this: "We use [Microsoft 365 or Google Workspace]. Which AI tool should we buy?"

A neutral consultant asks follow-up questions first. What do people spend their days on? How many work at a desk? What's already licensed? Then they give a reasoned answer that might be "use the Gemini you already pay for" or "license Copilot for 12 people, not 60."

A vendor-tied consultant answers before asking anything, and the answer is always the same product.

This is the single most useful step in the process. Ask every finalist for a 15-minute walkthrough of a real automation or assistant they built. Watch for:

• Does it handle a messy input? Ask them to try an incomplete form or an odd question.
• What happens when it fails? Ask to see the error alert and the fix process.
• Who maintains it? Ask whether the client's staff can change it without the consultant.
• What did it save? Ask for the before-and-after numbers and how they were measured.

A firm that has done real implementation work answers these easily. A firm that mostly writes strategy will struggle.

Before you sign, make sure the contract covers:

• Deliverables and acceptance. What "done" means, in writing.
• Ownership. You own all automations, assistants, prompts, documentation and accounts.
• Data handling. Which tools touch your data, confidentiality obligations and what happens to copies when the engagement ends.
• Licenses. Paid directly by you, or itemized at cost if the consultant buys them.
• Change orders. Scope changes priced and approved in writing before work.
• Support window. How long fixes are included after handoff.
• Exit. Notice period, handoff obligations and no fees to take your own work.
• Subcontractors. Who they are, at least by role, and that the primary firm stays accountable.

If you're comparing several proposals, put these terms in a simple table with one column per firm. Gaps jump out fast.

A national firm can do good work remotely. But for most small and mid-size Kansas City businesses, local has real advantages: on-site workshops, interviews in person, a consultant who understands the local labor market and industries, and someone who can come to your office when a rollout gets stuck.

The test isn't the address on the website. Ask where the people doing the work are based and whether they'll be in your building.

The safest way to choose is to test. Start with something contained: a readiness assessment, a policy package or a single workshop. You'll see how the consultant works, how they communicate and whether their recommendations make sense, before committing to a larger build. Our AI readiness assessment is built for exactly that. Prices are on our pricing page.

For what each kind of engagement typically costs, read what an AI consultant costs.

Book a Free AI Assessment Call ↗

We'd rather be picked for the right reasons. Book a free 30-minute call and run us through the list.

Book a Free AI Assessment Call ↗