How to choose an AI consultant in the UK

The short answer. Choose an AI consultant who starts by measuring your work, not by recommending a tool. A good consultant can tell you who will build what they recommend, show results from past projects in hours or pounds, explain how your data will be handled, and leave you owning everything they build. If they can't answer those four things clearly, keep looking.

What an AI consultant should actually do for you

For a small or mid-sized business, an AI consultant has one job: find the places where AI will save real time or money, then make sure the change actually happens. That usually means three things.

• Diagnose. Find out where your team's hours go today, and which of those tasks AI can take on reliably.
• Prioritise. Rank the opportunities by what they'd save against what they'd cost to build and run.
• Deliver. Build the automation, or manage whoever does, and measure the result.

Some consultancies only do the first two. That can work, but it leaves you finding and managing a builder yourself. We cover the difference in AI consultant vs AI agency vs doing it yourself.

Ten questions to ask any AI consultant

1. "How will you find out where AI fits in our business?"

You want to hear about time with your team: shadowing the work, mapping workflows, measuring how long tasks take. Be wary of anyone who already knows the answer before they've seen how you work.

2. "Who will build what you recommend?"

If it's someone else, find out who, and what happens when the build doesn't match the plan. Recommendations lose detail every time they change hands.

3. "What did your last two projects save?"

Ask for numbers: hours a week, pounds a year, response times. A consultant who measures their work will have them. "The client was really happy" isn't a result.

4. "Can we speak to someone you've worked with?"

Not every client agrees to be quoted publicly, but most consultants can arrange a call with one. If they can't, ask for a detailed case study instead.

5. "Which AI tools and models would you use, and why?"

A good answer depends on your systems, your data and your budget. Watch out for a consultant who recommends the same platform for every client, especially if they're a reseller for it.

6. "How will our data be handled?"

You need to know which data the AI will see, where it's processed, whether it's used to train anyone's model, and who can access it. UK GDPR still applies. Ask whether they hold a security certification such as Cyber Essentials Plus.

7. "Who owns what you build?"

The answer should be you. Automations should run on your accounts, with documentation your team can follow. Avoid setups where switching consultant means starting again.

8. "What will it cost to run once you've gone?"

AI tools have running costs: subscriptions, usage charges, hosting. Ask for a monthly estimate before you agree to the build. Our AI triage agent, for example, costs about £30 a month to run. Ask for that kind of figure up front.

9. "How will we know it's working?"

Agree the measure before the work starts. If the goal is saving time, someone needs to record how long the task takes now, so there's something to compare against later.

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

An honest consultant will tell you some things aren't worth automating yet. If everything they look at is a good fit for AI, they're selling, not advising.

Six warning signs

• They lead with the tool, not the problem. "You need Copilot" before they've asked how your team works.
• No measurement. No plan to record the before, so there's no way to prove the after.
• Vague deliverables. "AI strategy" or "transformation roadmap" with nothing concrete you can use on Monday morning.
• Lock-in. Automations that only run on their platform or their accounts.
• Hype words. "Revolutionise", "unlock", "game-changing". Good consultants say what the AI will actually do.
• No questions about risk. Nobody asks what happens when the AI gets something wrong, or who checks its work.

How to compare proposals fairly

Proposals from different AI consultancies rarely line up neatly. Put each one into the same five columns:

1. The problem it solves, in your words, not theirs.
2. What you get: the specific things you'll have at the end.
3. The total cost: the project fee plus a year of running costs.
4. The expected saving, in hours or pounds, and how it will be measured.
5. What happens afterwards: who supports it, and what that costs.

The cheapest proposal is often the one that leaves the most work for you to do afterwards. Our guide to what an AI consultant costs in the UK explains the common pricing models and how to work out whether a project will pay back.

Local or remote?

Most AI consulting can be done remotely. The exception is the discovery stage. Sitting next to the person who handles the inbox teaches you more in an hour than a week of video calls. If you can, choose someone who can be on site for the first few days.

Where Arx Certa fits

We're an AI consultancy for UK SMEs, based in Manchester. We're engineers, so the people who map your workflows are the people who build the automations. Every project is quoted in writing before it starts, you own everything we build, and we report the before and after in hours and pounds. Hold us to the same ten questions.

Frequently asked

Do I need an AI consultant, or can I do it myself?

If you have one clear task and someone on the team who enjoys setting up software, you can often start yourself with off-the-shelf tools. A consultant earns their fee when you don't know where to start, when the work spans several systems, or when the data involved is sensitive.

What's the difference between an AI consultant and an AI consultancy?

Usually just size. A consultant is often one person. A consultancy is a firm, which may cover more skills, such as security, data and software development. What matters more is whether the people advising you can also deliver.

What qualifications should an AI consultant have?

There's no single recognised qualification for AI consulting. Look instead for evidence of delivery: projects with measured results, a technical background in software or data, and a security certification such as Cyber Essentials Plus if they'll be handling your systems.

How long does an AI consulting project take?

A discovery or readiness assessment typically takes a couple of weeks. A first automation build usually takes a few weeks more. Be cautious of long strategy phases with nothing built at the end.