Have you just been quoted somewhere between five and twenty thousand pounds for an “AI system” and you’re still not entirely sure what you’d actually be getting for the money?
You’re not alone, and you’re not being dim. Most of these quotes are genuinely hard to pin down, and that’s not always by accident.
We’ve sat in on quite a few of these conversations recently, either helping a client make sense of a proposal they’ve already been sent, or building something ourselves once a client has worked out what they actually need. And the pattern we keep seeing is the same one, over and over.
Somebody watches a slick demo. The AI answers questions, drafts an email, sorts some data, whatever the trick of the day happens to be. It looks brilliant. Fifteen minutes later there’s a proposal in their inbox with a number on it that would make most business owners wince.
What often doesn’t happen is anybody explaining, in plain English, what is actually being built, how long it will take to work properly, and what happens when it doesn’t.
This article isn’t about whether you should use AI in your business at all. We’ve written separately about that. This one assumes you’ve already decided to look into it and you’re now staring at a quote, trying to work out whether it’s fair, sensible, or a fairly polite way of relieving you of some money.
Before We Go Any Further
This is simply our view, based on what we’ve seen working with businesses over the years.
We’ve watched brilliant systems get built for a fraction of what businesses expected to pay. We’ve also watched businesses spend five figures on something that, six months later, nobody in the company was actually using.
Have a look around, ask other business owners what they’ve experienced, and make your own judgement. What follows is simply how we currently see things.
Why Do AI System Quotes Vary So Wildly?
Because “AI system” doesn’t actually describe a fixed thing. It’s a bit like asking someone to quote you for “a website” without saying whether you mean a single landing page or a fully custom booking platform with five hundred pages of content behind it.
An AI system could mean:
- A chatbot that answers common questions from your website copy.
- A tool that reads incoming documents and pulls out the relevant data.
- Something connected to your CRM that drafts follow-up emails.
- A voice system that answers your phones.
- A genuinely custom-built workflow that touches multiple parts of the business.
Those are wildly different jobs, with wildly different costs, timescales and levels of ongoing maintenance. If a supplier hasn’t clearly told you which one of these (or something else entirely) you’re actually buying, that’s the first thing to sort out before you talk about price at all.
What’s the Difference Between a Demo and a Working System?
Roughly the same difference as between a car that starts in the showroom and a car that gets you through a British winter for five years without letting you down.
A demo is built to work once, under ideal conditions, in front of an audience who wants to be impressed. That’s fine. Demos are meant to sell an idea.
A working system has to cope with the messy reality of your actual business. Customers who phrase things oddly. Documents that don’t quite match the format it was trained on. Staff who use it slightly differently to how it was designed. Edge cases nobody thought to test.
“A demo tells you the idea works. It doesn’t tell you the system works. Those are two very different questions, and only one of them should get your signature.”
The gap between the two is where most of the real cost, and most of the real risk, actually sits. It’s also usually the bit that gets glossed over quickest in a sales conversation, because it’s the least exciting part to talk about.
What Does “AI-Powered” Actually Mean on Your Invoice?
Often, not very much. Sometimes it means a genuinely sophisticated piece of engineering. Sometimes it means somebody has wired up an existing AI model to your website and added a markup that would make a jeweller blush.
To be fair, some of the prices we’re seeing businesses charged for relatively simple AI setups are ridiculous. We’ve reached the point where adding “AI-powered” to something sometimes appears to add another zero to the quotation, without necessarily adding proportionate value.
That doesn’t mean every AI product is overpriced. Some genuinely require significant custom development, ongoing training, integration work and support, and the price reflects real effort. Our point isn’t that AI should be cheap. Our point is that the price needs to relate to what’s actually being built, and you’re entitled to understand that relationship before you agree to it.
If somebody wants £15,000 from you but cannot explain in plain English what they are actually building, that’s a fairly big warning sign for us.
Imagine a Business That Gets Quoted £10,000 for a “Custom AI Assistant”
We don’t have a real case file to hand you here, so let’s be clear this is a hypothetical, but it’s a fairly typical shape of thing we’ve seen play out.
Say a small trades business gets quoted £10,000 for a “custom AI assistant” that handles enquiries and books jobs into the diary. It sounds fantastic. The demo, shown on the supplier’s own test data, works beautifully.
Three months and £10,000 later, the business discovers the “assistant” is largely a pre-built chatbot template, lightly customised, sitting on top of a subscription the supplier is also quietly marking up. Nobody explained, before the contract was signed, what would happen if the AI misbooked a job, or what training would be needed for it to actually understand the business’s real services, pricing and availability.
The lesson isn’t “never buy an AI assistant.” It’s that if the business had asked a handful of direct questions upfront, about what was custom-built versus off-the-shelf, who owned the underlying data, and what happened when something went wrong, they’d have gone in with realistic expectations, and probably negotiated a fairer price.
What Questions Should You Be Asking Before You Sign?
Start with this one: if this supplier disappeared tomorrow, what would you actually be left with?
That single question tends to expose a lot. If the honest answer is “not much, because it all lives on their infrastructure under their account,” you need to know that before you pay, not after.
From there, ask what’s genuinely custom versus what’s a re-skinned existing product. Ask who owns your data once it’s fed into the system. Ask what training or setup time is included, and what happens once that runs out. Ask what “support” actually covers, and whether it’s included or billed separately.
None of these questions should make a decent supplier uncomfortable. If they do, that tells you something too.
“If someone can’t explain in plain English what their system is actually going to do for your business, I’d think very carefully before handing them thousands of pounds.”
So, What Should You Actually Do?
Before you sign anything, slow the conversation down. A good supplier won’t mind. A pushy one might, and that’s useful information in itself.
- Get the scope written down in plain English, not jargon.
- Ask exactly what’s custom-built versus off-the-shelf.
- Find out what it costs monthly, not just upfront.
- Confirm who owns the data and where it’s processed.
- Ask what happens if the AI gets something wrong.
- Check whether a human can override or step in.
- Ask what you’re left with if you cancel or the supplier folds.
- Get a clear view on what “support” actually includes.
- Ask how you’ll measure whether it’s actually working.
- Get at least one other quote before you commit.
Before Buying an AI System, Ask:
- What problem does this actually solve, in plain terms?
- What is genuinely custom, and what is off-the-shelf with a markup?
- What will it cost to build, and what will it cost every month after that?
- What information does it need access to, and where is that information stored?
- Who owns the data once it’s inside the system?
- What happens if it gets something wrong, and who’s accountable for fixing it?
- Can a human step in and override it when needed?
- Can you export your own data if you want to leave?
- What happens to the system if the supplier stops trading?
- How will you actually measure whether it’s working?
If a supplier can answer all of those clearly and confidently, that’s a genuinely good sign. If they get vague, defensive or start talking about “proprietary technology” instead of answering the question, slow down.
Frequently Asked Questions
How much should a small business expect to pay for an AI system?
There’s no fixed answer, because “AI system” covers everything from a simple chatbot to a fully custom integration touching multiple parts of the business. The price should relate directly to what’s genuinely being built. Get the scope in writing before you talk about cost.
Is a demo a reliable way to judge whether an AI system will actually work?
Not on its own. A demo shows the idea works under ideal conditions. It doesn’t show how the system copes with your real customers, your real data and your real edge cases. Ask to see it tested against something closer to your actual business.
What does “AI-powered” actually mean?
It can mean genuinely sophisticated custom engineering, or it can mean an existing AI model with a wrapper and a markup on top. The phrase itself tells you very little. Ask what’s actually being built underneath it.
Should I be worried if a supplier can’t explain their system in plain English?
Generally, yes. If someone genuinely understands what they’re building, they should be able to explain it to you without hiding behind jargon. Struggling to explain it plainly is often a sign that the explanation doesn’t hold up well under questioning.
Is it reasonable to ask what happens if the supplier goes out of business?
Completely reasonable, and worth asking every time. You want to know whether your data and any custom work are portable, or whether they disappear along with the supplier.
Are all expensive AI systems overpriced?
No. Some genuinely complex, well-built systems are worth every penny of a five-figure price tag. Our point isn’t that AI should be cheap. It’s that the price should relate to what’s actually being delivered, and you’re entitled to understand that before you commit.
Our Final View
AI can be genuinely useful, and there are suppliers out there doing excellent, honest work.
There are also suppliers relying on the fact that most business owners don’t yet know the right questions to ask, and pricing accordingly.
Our view is fairly simple. Get the scope in plain English. Understand what’s custom and what isn’t. Ask what happens when things go wrong, and what you’re left with if the relationship ends. Then decide whether the price actually matches what’s being built.
You don’t need to become a technical expert to buy an AI system sensibly. You just need to ask a few direct questions and expect direct answers.
Looking at an AI Quote and Not Sure What You’re Actually Being Sold?
If you’ve had a proposal land in your inbox and you’re not entirely sure whether it stacks up, give us a call and talk us through it.
You don’t need another sales pitch. Sometimes it just helps to run it past somebody who’s seen a few of these before you sign anything.
Pick up the phone. We’re always happy to talk it through.