How to Choose an AI Consultant for Your Small Business: 10 Questions to Ask

A year ago, "AI consultant" was a job title almost nobody had. Now every web agency, marketing freelancer and software reseller offers it. Some of them are excellent. Some are selling a chatbot template with your logo on it and a contract you cannot easily leave.

You do not need to understand how large language models work to tell the difference. You need to ask the right questions and listen carefully to the answers. Here are the ten we would ask if we were on your side of the table.

1. "What job would you start with?"

A good AI consultant wants to talk about your work before they talk about technology. They will ask what eats your team's week, where things get dropped, and which tasks people dread. Their first suggestion should be one specific, repetitive job, such as answering the twenty questions customers ask most, pulling data out of supplier invoices, or drafting replies for someone to approve.

Red flag: they open with a platform, a tool or a "full AI transformation" before they know how your business runs.

2. "Do I actually need AI for this?"

This is the most revealing question on the list. A lot of what businesses call "AI" is really rules: when a booking comes in, send a confirmation, create an invoice, add the customer to the mailing list. That is plain automation, and it is cheaper, faster and more predictable than a model.

An honest consultant will tell you when you do not need AI. If you are not sure where the line sits, we wrote a plain-English guide to automation vs AI.

Red flag: every problem you describe gets the same answer, and the answer is AI.

3. "What data will it need, and where does that data go?"

AI that answers customer questions or reads documents has to work from something: your website, policies, price lists, past emails, PDFs. Ask which of your data the system will use, which third-party services it passes through, and how long it is kept.

You should get a clear, specific answer. If you work in a sector with confidentiality obligations, the consultant should be designing around them from day one, not discovering them after launch.

4. "How will it avoid making things up?"

A customer-facing assistant that invents an answer is worse than no assistant at all. Ask how the system stays grounded in your own material, and what happens when it does not know.

The answer you want sounds like: it answers from your documents, it says "I don't know, let me pass you to the team" rather than guessing, and anything high-stakes goes to a human first.

5. "Where does a human stay in the loop?"

Good AI projects are rarely fully hands-off on day one. The usual pattern is draft and approve: the system writes the reply, fills in the form or flags the invoice, and a person checks it with one click. You still save most of the time, and nothing embarrassing reaches a customer.

Ask where the consultant would keep a human involved, and how you would decide to remove that step later.

6. "How will we know it works?"

"It worked in the demo" is not a test. Ask how they will measure accuracy before it goes live. A serious answer involves a test set: a batch of real examples from your business (real enquiries, real invoices) with the correct answers, run through the system before launch and again after every change.

Red flag: success is described in terms of how impressive it looks rather than how often it is right.

7. "What will it cost to run each month?"

AI has a running cost that ordinary software does not: you pay for usage. A consultant should estimate this during scoping, explain what makes it go up, and use cheaper models for the parts of the job that do not need the expensive ones.

If nobody mentions running costs until your first bill arrives, that tells you how the rest of the project will go.

8. "Who owns what you build?"

This is where small businesses get caught out most often. Ask directly: - Who owns the accounts the system runs on: you or the consultant? - Can you see, export and change the prompts, workflows and data? - If you stop working together, what keeps running and what switches off?

The right answer is that you own your accounts, your data and your workflows, and the consultant works inside them. You should be able to hand the whole thing to someone else without starting again.

9. "Will you show us how it works?"

Some agencies make themselves permanently necessary by keeping everything a black box. Better ones document what they build and teach your team to manage the day-to-day: updating the information the assistant answers from, changing a template, pausing a workflow.

We feel strongly about this one. As one of our clients put it in a review, the thing that stood out was that we were not "trying to trap me into relying on their services forever." You might still want ongoing support, and that is fine. It should be a choice, not a hostage situation.

10. "Can we start small?"

A sensible first engagement is narrow: one job, scoped, built, tested and live within weeks. If it works, you expand. If it does not, you have learned something useful without betting the business on it.

Red flag: a long minimum contract before anything has proved its value.

What a good first conversation looks like

By the end of a first call with a good AI consultant, you should have: - A shortlist of two or three jobs worth automating, ranked by time saved - An honest view of which need AI and which only need automation - A rough idea of the data involved and the risks - A clear next step that does not require you to sign anything long-term

If you want a head start before that conversation, our free guide The 20 Jobs You Should Automate First lists the tasks we see paying back fastest in small businesses.

Working with Digital Rogues

We are a small team of AI and automation consultants who also build websites, which means we see where the time actually goes in a business. Every project starts with one job and a clear measure of success, and you keep ownership of everything we build. If you would like an honest opinion on where AI fits in your business (or whether it does), take a look at our custom AI work or get in touch.