What AI costs a small UK business, measured against one job

“How much does AI cost” is the wrong question, and it is why so many small businesses either overspend or never start. A tool is not expensive or cheap in the abstract. It is expensive or cheap against one job.

For a domestic electrician with an average job around £250, almost every AI subscription on the market pays for itself if it recovers a single job a quarter. That is the number that should decide it, and almost nobody works it out before buying.

The arithmetic that actually decides it

Take the situation from the one-person trade example: the phone rings four or five times a day while you are on a job. Say one of those a day is genuinely new work, and your average job is £250.

1 new enquiry per working day
× 5 days
= 5 new enquiries a week

5 × £250 = £1,250 of new work passing through the phone each week

You do not lose all of it. Some callers leave a message, some ring back, some you catch between jobs. But you have no idea what share you lose, because a caller who rings the next electrician never tells you they did.

That uncertainty is the point. If you cannot see the loss, you will not price a solution against it, and you will judge a £30 a month tool on whether £30 feels like a lot rather than on whether it recovers £250.

The break-even, in one line

Monthly cost ÷ your average job value = jobs per month to break even

At £250 a job:

Monthly cost	Jobs a month to break even
£20	0.08, roughly one job a year
£50	0.2, one job every five months
£100	0.4, one job every ten weeks
£300	1.2, more than one a month

Under £100 a month, a tool that recovers one job a quarter is already ahead. Above about £300 a month you are committing to it producing work every month, which is a different kind of decision and needs evidence first.

Do this with your own numbers before you look at any product. A business with an £80 average order has completely different maths, and the same tool that is obvious for the electrician is marginal for a café.

What the price bands actually buy

Rather than quote figures that move, the shape of the market as of August 2026, with the vendor’s own pricing page linked so you check rather than trust:

Free tiers. Enough to test whether a task is automatable at all. Not enough for anything with customer data in it, because the terms on consumer tiers usually differ from business tiers on training and retention. Fine for two weeks of proving the idea.

Around £15 to £30 per person per month. General assistants on a business plan: ChatGPT, Claude, Microsoft 365 Copilot, Google Workspace. This band covers quotes, drafting, summarising and admin. For a one or two person business this is usually the whole AI budget and it is enough.

Around £30 to £100 a month. Job-specific services: AI call answering, booking, review chasing. Priced per user, per minute or per call, and the per-minute ones need a usage estimate before you sign anything.

£100 a month and up. Industry software with AI attached, usually field service or trade management platforms. Judge these on the software, not the AI. You are buying scheduling and invoicing that happens to have AI features.

Always check whether UK pricing and VAT are included, and whether the headline figure is annual-billed. The monthly-billed price is frequently 20 to 25% higher, and for a first trial the monthly price is the one you want.

The cost that is not on the pricing page

The subscription is usually the smallest number in the decision.

Setup time. Two to six hours for most tools in this band, and it is your time or a staff member’s. At any realistic hourly rate that is often more than the first year of licence fees.

Checking time. Every AI output that touches a customer needs a human look until you trust it. If checking takes longer than doing, the tool has a negative return no matter what it costs. This is the failure that kills most first attempts and it never appears in a pricing comparison.

The switching cost of getting it wrong. Annual plans, data in a proprietary format, a phone number ported to a service you want to leave. Start monthly.

Nothing at all, if it does not work. The most common real outcome of a £20 a month trial is a decision, which is worth the £40 you spent over two months.

How to decide, in fifteen minutes

1. Write down your average job or order value. The real one, from your invoices, not the one you quote people.
2. Divide the tool’s monthly cost by it. That is your break-even in jobs.
3. Ask whether that number is plausible. One job a quarter, usually yes. Two jobs a month, you need evidence first.
4. Add the setup hours at your own rate. If that changes the answer, the tool is more expensive than it looks.
5. Start monthly and set a review date. Two weeks to decide, per where to start with AI.

What not to do

Do not buy annual to save 20%. You are betting a year of fees against a tool you have not tested. The saving is real and the bet is bad.

Do not price against your salary. “It costs less than an hour of my time” is true of almost everything and justifies almost everything. Price against a job, which is a number the business actually earns.

Do not add tools without removing any. Three £25 subscriptions is £900 a year, and small businesses accumulate these without ever running the subtraction.

Do not skip the data question because the price is small. A £20 tool handling customer addresses carries exactly the same obligations as a £2,000 one. Twenty minutes on whether you need a DPIA covers it.

---

Prices in this band move constantly. Every figure above is a band rather than a quote, and every link goes to the vendor’s own pricing page. Check the number on the day you buy, not the day you read this.