The short version: Before you hand anyone money to "do AI" for your business, make them answer questions about their own tools, their pricing, their failures, and what happens in week one, not just their vision for your future. If they get vague or defensive on any of it, that's your answer already.
Why I started keeping this list
A client of mine, a manufacturing company in the Midlands with about 40 staff, called me in 2024 in a proper state. They'd paid a consultant £4,000 for what he called an "AI transformation roadmap." It was a 30-page PowerPoint deck full of buzzwords like transformation and future-readiness, with no actual tools named, no cost breakdown, no plan for who would build anything, and not one line about their existing systems. The owner asked him one direct question in the follow-up call, "what do we do on Monday morning," and the man couldn't answer it.
That deck cost more than most small businesses spend on their first three months of real AI implementation. I know because I've quoted what a proper 90-day AI implementation costs and delivers, and it's not a slide deck. It's working automations, trained staff, and measured time saved.
Since then I've built a short list of questions I tell every business owner to ask before they sign anything. Not because I want to scare people off hiring help. I run this exact service. It's because the AI consulting space right now has almost no barrier to entry, and that's the uncomfortable bit nobody selling you their services wants to say out loud.
The uncomfortable truth about the job title "AI consultant"
There is no licence for this. No exam board, no professional body, no minimum years of experience required. Someone can do a weekend course on prompting, print business cards, and call themselves an AI consultant by Monday. I've met people doing exactly that, charging £150 an hour to run a two-hour ChatGPT demo they learned from a YouTube video the week before.
That doesn't mean every consultant is a chancer. Some of the best people I know in this space have zero formal AI qualification and are brilliant, because they've built and broken things in real businesses. But it does mean the title itself tells you nothing. You have to interrogate the person, not the label. That's what the rest of this list is for.
Start with money, not magic
Ask this before anything else: "What exactly will I have at the end of this, and what does it cost, broken down?"
- A good answer names specific deliverables: three automated workflows, a trained inbox response system, a documented process handover, staff training sessions, and a follow-up check at 30 and 60 days.
- A bad answer says things like "clarity," "alignment," or "strategic direction" and no number.
On UK pricing specifically, expect day rates somewhere between £600 and £1,800 depending on seniority and location, and project-based engagements for a small business running from roughly £2,500 for a narrow single-process build up to £15,000 to £20,000 for a full 90-day implementation across several departments. Retainers for ongoing support usually sit between £1,000 and £4,000 a month. If someone quotes you £30,000 for a first project with no phased milestones, or £300 for "full AI transformation," both should worry you.
1. "Can you show me something you built for a business like mine, and can I speak to that client?"
Not a case study PDF with a logo and a made-up quote. An actual person you can call. I give every prospective client at least two references from the last six months, and I expect to be asked. If a consultant hesitates or says references are "confidential," push. Confidentiality about strategy is normal. Confidentiality about whether the work worked is a dodge.
2. "What tools do you use day to day, and why those ones?"
This question separates people who work in AI from people who talk about AI. I want to hear specific names, ChatGPT Enterprise versus the free version and why that matters for data handling, Zapier or Make for connecting systems, Claude for longer document work, a specific CRM's native AI features rather than a bolt-on nobody's heard of. If they answer in vague terms like "the latest AI models" with no names, they may not be hands-on themselves.
3. "What's the first thing you'd automate in my business, and how do you know?"
They shouldn't be able to answer this in detail on a first call, and that's fine, they don't know your business yet. But they should be able to describe their process for finding out. Do they ask for a week of your team's task list? Do they shadow a call handler? Do they look at your ticket volume? If their process is "we'll have a workshop," ask what happens in the workshop, specifically. A vague process produces a vague plan. I've seen this go wrong industry by industry, which is part of why I wrote up what to automate first across 19 different sectors, because the right starting point for an estate agent is nothing like the right starting point for an ecommerce store.
4. "What happens if it doesn't work?"
Ask this one directly and watch their face. A consultant who's confident in their own delivery will have a real answer, phased payment tied to milestones, a get-out clause after 30 days, or a written statement of what "success" looks like before you start so failure is measurable, not a feeling. Someone who's never had a project fail, or won't admit they have, is either new or not telling you the truth. I've had projects underperform against the original target. I say so when it happens, and I explain what we changed. That's normal. Pretending it never happens isn't.
5. "What data of mine will you need, and where does it go?"
This is the question most business owners skip, and it's the one that can hurt you most. If a consultant wants to feed your customer database, contracts, or staff records into a general AI tool without a data processing agreement, or without knowing your obligations under UK GDPR, stop right there. Ask specifically whether they use enterprise-tier tools with data protection built in, whether anything you share trains a public model, and whether they'll sign an NDA and a data processing agreement before they see anything sensitive. A consultant working with ecommerce customer service data, order histories, refund patterns, personal details, has no excuse for being loose about this.
6. "How will you train my team, not just build the thing?"
This is where a lot of AI projects die quietly six months later. The consultant builds a slick automation, everyone's impressed on demo day, then the one staff member who understood it leaves and nobody else can touch it. Ask exactly how many training sessions are included, whether there's written documentation your team keeps, and whether they'll come back for a check-in after the initial rollout. I build this into every 90-day engagement because a tool nobody in your business can maintain isn't a deliverable, it's a liability with a nice interface.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
7. "What have you built with AI for your own business?"
I love this question because it's uncomfortable for the wrong people to answer. If someone is telling you to automate your customer enquiries, your content, or your admin, but their own business runs entirely on manual spreadsheets and a personal assistant doing everything by hand, ask why. There's nothing wrong with a consultant who's selective about what they automate in their own operation, plenty of good reasons exist. But they should be able to tell you what they use and what they've deliberately chosen not to, with reasoning, not silence.
8. "How quickly can you show me a result, and what does 'result' mean here?"
Push for a number and a date. "Within four weeks you'll see a documented reduction in average response time on customer enquiries, measured against your current baseline." That's a real answer. "You'll start seeing benefits fairly quickly" is not. If you run an estate agency, for example, the fastest payback tasks are usually specific and known in advance, not a mystery to be discovered mid-project, which is exactly why I laid out which tasks pay back fastest for estate agents rather than leaving it vague.
Red flags worth walking away from
- They can't name a single AI project that went wrong for a past client.
- Their pricing is a single flat number with no breakdown by phase or deliverable.
- They talk exclusively in the future tense, "AI will transform your industry," and never in the specific present tense about your business.
- They can't explain, in plain English, what a large language model does or doesn't do well. If you ask "can this AI make decisions on its own about refunds" and they say yes without qualification, that's wrong, and it's a sign they don't understand the limits of what they're selling you.
- They push you toward the most expensive package before they've asked a single question about your current processes.
- No written contract, no data agreement, just a handshake and an invoice.
What to ask if you're a very small business or solo operator
If you're a one or two person business, the questions shrink but they don't disappear. Ask how much of the work you'll be able to run yourself after they leave, because you likely can't afford an ongoing retainer. Ask for the cheapest version of the fix that still works, not the most impressive one. And ask them directly whether your business is even ready for this yet, because sometimes the honest answer is that your processes need tidying before any AI tool will help, and a consultant who says that out loud, even though it costs them the sale, is one worth trusting. I cover the practical side of this in more detail in how to hire an AI consultant for a small business, including what a fair first project looks like at that size.
A quick sanity check before you sign
Once you've asked all of the above, do one more thing. Ask them to explain your own project back to you in three sentences, in plain English, with no jargon. If they can't, either they haven't listened to your business, or they're relying on jargon to hide the fact there isn't much substance underneath it. Either way, that's information you needed before you paid, not after. If you want a fuller sense of what fair pricing looks like across different project sizes before you get quotes, it's worth reading through what an AI consultant costs so you're negotiating from a position of knowing the market, not guessing.
If this is the kind of work you want done rather than read about, start with how to hire an AI consultant, then see what I do as an AI automation consultant.
A closely related guide: what working with me looks like.
Frequently asked questions
What is the single most important question to ask an AI consultant before hiring them?
Ask what exactly you'll have at the end of the project and what it costs, broken down by deliverable. A consultant who can't give specifics here, and answers only with vague words like transformation or strategy, hasn't planned real work, they've planned a sales pitch.
How much should an AI consultant cost in the UK?
Day rates typically run from around £600 to £1,800 depending on experience and location, project-based work for a small business usually falls between £2,500 and £20,000 depending on scope, and ongoing retainers tend to sit between £1,000 and £4,000 a month. A quote far outside those ranges, in either direction, deserves a direct question about what's included.
Do AI consultants need a qualification or certification?
No. There's no licensing body or required qualification for the title AI consultant in the UK or anywhere else, which means anyone can use it after a short course. This makes references, past client conversations, and specific evidence of hands-on tool use far more important than any certificate they show you.
What's a fair timeline to expect results from an AI consulting project?
For a focused, well-scoped project, most small businesses should see a measurable result, a documented time saving or response time improvement, within four to six weeks of starting, with a fuller rollout across multiple processes complete by around 90 days. If a consultant can't give you any timeline at all, that's a sign the project hasn't been scoped.
