AI just got affordable for small businesses: but cheap tools aren't the same as a working system

If you looked at AI for your business 18 months ago and decided it was too expensive, look again. The cost of running the most useful jobs has fallen sharply. Things that needed a developer and a £10,000 build at the start of 2024 now run on a £20 to £80 monthly subscription. Most owners have not caught up with how far the running cost has already fallen.

This is the post for the owner who took one look at AI in 2024, saw the price tag, and quietly decided to wait. The wait is over. But the cheap tools are not the working system, and that is the part nobody tells you in a launch tweet. The honest answer is that picking the right tool for your business and wiring it into the way you already work is the part that pays back, and that part has not got cheaper at all.

This piece walks through three SME jobs that now cost less than £100 a month each on the tooling, the typical month's payback for each one, and the typical reason most owners who DIY them never see the payback show up.

For the wider pricing view, our AI automation pricing guide and our AI chatbot cost guide cover the build-side numbers underneath this affordability piece.

Three jobs that now cost under £100 a month

Same shape every time: a job that used to be too expensive to consider, that the price drop has put inside reach for an owner-led business.

Job one: answering customer enquiries round the clock. A small AI agent that watches your phone number, your website chat, and your social DMs. Answers FAQs, qualifies the enquiry, takes a booking or hands off to you. Tooling cost: roughly £40 to £80 a month for the model and the messaging layer. Typical payback: the first month for any business that loses bookings to unanswered phones or out-of-hours messages. A salon that recovers two missed bookings a month at £80 each is already ahead. A restaurant that captures one Friday booking from a Tuesday-evening DM is already ahead.

Job two: chasing missed appointments and getting them rebooked. An automation that watches your booking system, picks up the no-shows or the cancellations, and sends a personalised text or message offering a rebook within a fixed window. Tooling cost: roughly £20 to £60 a month for the model and the messaging. Typical payback: the first week for any business with a no-show rate above 10%. A clinic that rebooks four appointments a month at £100 each is in clear profit on the tooling.

Job three: generating social posts and first-draft review replies. A weekly AI drafter that turns your news (a new menu, a new treatment, a recent review) into a week's worth of social posts in your voice, plus a first-draft reply to every Google review the moment it lands so you can polish and post in 30 seconds rather than 30 minutes. Tooling cost: roughly £20 to £40 a month. Typical payback: time saved rather than new revenue, but for an owner who currently writes their own social on a Sunday evening, the hours saved are real.

Three jobs, under £300 a month between them on the tooling alone. Eighteen months ago, the build for these jobs would have cost more than that as a one-off, and the running cost would have been a similar number every month again.

Why the DIY version rarely pays back

If the tools are this cheap, why does the average owner who buys them not see the payback? Three honest answers, the same ones we hear in nearly every conversation.

One: the wrong tool for the business. There are dozens of small AI assistants on the market. Most of them work for somebody. Few of them work for you specifically. The owner who picks the cheapest one in the search results, plugs it in, and waits for the bookings to roll in usually finds the tool was built for an industry next to theirs, with handoffs that do not match how their business actually runs. Two weeks later, the trial is up and nothing has changed.

Two: the wiring is missing. A working AI workflow is the model plus the wiring. The wiring is the part that hands the customer back to your booking system, sends the message from the right phone number, logs the conversation against the right customer record, and tells you when something needed your attention. Without the wiring, the AI agent answers a question in a tab nobody is watching. The owner gets the right answer; the customer never sees it.

Three: the wording is wrong. Default settings give you a model that sounds like a model. The customer who calls your salon expects a salon voice, not an assistant voice. The owner who never sets the wording, the offer policy, or the handoff rules ends up with messages that customers find awkward, and the owner turns the system off after the third complaint. The fix is short (a paragraph of brand voice, a clear offer policy, a defined handoff rule) but if it is missing, the system loses trust faster than it builds revenue.

What the working version involves

The working version of any of these three jobs has four parts.

One. A short conversation about what your business actually does, where the bookings and revenue come from, and which channels are leaking. This is what the AI Opportunity Audit covers in 30 minutes.

Two. A tool choice that fits your business shape. Not the cheapest, not the most marketed, the one whose handoffs and pricing match your business.

Three. The wiring: connecting the AI to your booking system, your phone number, your inbox, and your customer record so the result lands in the right place automatically.

Four. The wording: a paragraph of brand voice, a clear offer policy, a defined human handoff rule. Reviewed quarterly as the business evolves.

The tooling cost is the part you see on a card statement. The four parts above are the part that turns the tooling cost into the result. We sell those four parts. The tools we recommend are whatever fits the business; we do not resell software, we build the system on top.

The honest answer

If you have been waiting for AI to get cheap, it has. The three jobs above are inside reach for any owner-led business in 2026 from a tooling perspective. If you are doing the work yourself, set aside a Sunday for the wording and a Saturday for the wiring on each one, and expect a few iterations before it settles. The savings will arrive when the system is dialled in, not the day you sign up.

If you would rather have it dialled in for you, that is what the implementation work is. The tools are cheap. Picking the right one and getting it working in your business is what we do.

When the cheap version is the right call

Not every business needs the implementation. Three honest signals that the DIY version is the right call for you.

One. You are an owner who genuinely enjoys the technical work and has spare weekend hours to spend on it. The tooling is friendly enough that a curious owner can wire up a single workflow over a weekend if the workflow shape is simple.

Two. The job you are trying to automate is a clean fit for one of the popular off-the-shelf assistants, with no awkward handoff to a booking system, a payment system, or a regulated process. A dog walker generating weekly newsletter posts is a different problem to a clinic chasing missed appointments.

Three. You can afford the time to run it for two months and switch it off if it does not pay back, without losing trust with customers in the meantime. Some workflows are forgiving on a bad first version (social posts), some are not (anything customers see in the moment).

If those three signals fit, run the DIY version. If even one of them does not, the implementation work usually pays back inside the first quarter. The tooling is the small line item; the result is the asset.

The useful question

The useful question is not whether AI is cheap now. It is whether there is a specific leak in your business where a cheap running cost would pay back once the workflow is wired in properly.

If you want a 30-minute conversation about where that first leak is, book a free audit. We will tell you whether the right answer is a simple DIY test, a consultant-built workflow, or waiting until the underlying process is cleaner first.