Ask a room of small business owners why they have not adopted AI and you will hear the same answer, delivered with quiet confidence. “It doesn’t really apply to what we do.”
It sounds reasonable. It is also, in most cases, wrong. And it is quietly becoming one of the more expensive assumptions in British business.
The number one barrier is not what you think
When researchers ask SMEs that have not adopted AI what is holding them back, the answers are revealing. Recent survey research found that 77% of non-adopters say there is no applicable use case for their business. That puts perceived irrelevance ahead of lack of understanding, cited by 62%, and having no in-house expertise, cited by 60%.
Read that again. The biggest barrier is not cost. It is not risk. It is not a shortage of skills, though that ranks high. The biggest barrier is the belief that AI simply has nothing to do with the business in question.
That belief deserves scrutiny, because it rarely survives contact with the evidence.
What the businesses on the other side are seeing
Look at the SMEs that did adopt, and the picture is very different. Salesforce research reports that 91% of small businesses using AI see revenue increases. Not marginal efficiency tweaks buried in a spreadsheet. Revenue.
The same Salesforce research shows where those businesses started. Customer service is the most common working entry point, used by 51% of adopters, followed closely by content creation and email. These are not exotic applications. They are the ordinary, unglamorous jobs that exist in almost every business: answering the same questions repeatedly, writing product descriptions, chasing invoices, drafting follow-ups.
Here is the uncomfortable implication. If 91% of adopters see revenue gains, and the most common use cases are things nearly every business does, then the 77% who see “no applicable use case” are not describing their businesses accurately. They are describing their view of their businesses.
“No use case” almost never means the use case does not exist. It means nobody has gone looking for it.
Why sensible people reach the wrong conclusion
The perceived irrelevance problem has a specific cause, and it is worth naming because it is fixable.
Most owners evaluate AI by imagining it doing their most important work. The judgement calls. The client relationships. The craft that makes the business what it is. AI looks obviously wrong for those jobs, so the conclusion follows: not for us.
But AI’s value in an SME almost never sits in the important work. It sits in the repetitive work that surrounds it. The important work is what you notice. The repetitive work is what you have stopped noticing, precisely because you do it every day.
An accountancy practice does not need AI to give tax advice. It needs AI to draft the forty near-identical client emails that go out every week. A trades business does not need AI to fit a boiler. It needs AI to turn a voicemail into a booked appointment without anyone retyping the details.
Nobody sees these use cases from the top of the business, because from the top they are invisible. They only appear when someone maps the work.
The five-times-a-week test
Here is the method we use, and it costs nothing but an hour and a sheet of paper.
List every task your team does more than five times a week. Not the strategic work. The routine: quotes, replies, data entry, status chasing, report formatting, appointment juggling, copying information from one place to another.
Then go through the list and mark three things.
Which tasks follow rules? If a person can write down the steps, or already follows an unwritten script, the task is a candidate for automation or AI assistance.
Which tasks produce text? Emails, quotes, descriptions, summaries, responses. Generating and drafting text is the thing current AI does best, and it is where the Salesforce entry points, customer service, content and email, all live.
Which tasks involve moving data between systems? Retyping an order from email into the job sheet. Copying invoice details into the accounts package. This is pure friction, and it is highly automatable.
The AI use case is almost always somewhere in that list. Usually it is in the first three items, staring back at you. The exercise does not require technical knowledge. It requires honesty about how much of the working week is spent on work a machine could draft, sort or shuttle.
The opposite mistake is just as expensive
Before anyone rushes off to buy licences, a warning. There is a second failure mode, and it is the mirror image of the first.
Plenty of businesses “adopt AI” by signing up for a chatbot subscription, using it sporadically for a few weeks, and letting it drift. No process was mapped. No task was chosen. No measurement was set. The tool floats free of the actual work, and after three months nobody can say what changed.
Tools without process mapping produce dabbling, and dabbling is what a lot of survey-reported adoption actually is. It inflates the statistics and deflates the results. It is also how businesses end up concluding, wrongly for a second time, that AI does not work for them. First they assumed it did not apply. Then they applied it to nothing in particular and proved themselves right.
The fix for both failures is the same: start from the work, not the tool.
What to do about it
If you suspect your business is in the 77%, here is the practical path.
Run the five-times-a-week exercise this month. One hour, whole team if you can. List the repetitive tasks, then mark which follow rules, which produce text, and which move data between systems.
Pick exactly one task. Choose the one that is high frequency, low risk and clearly bounded. Drafting customer service replies for human review is a strong first pick, which is precisely why 51% of adopters in the Salesforce research started there.
Define what success looks like before you start. Hours saved per week, response time, quotes sent per day. If you cannot measure it, you will end up dabbling.
Keep a person in the loop. For a first use case, AI should draft and a human should approve. You get the speed without betting the customer relationship on an unproven tool.
Give it six weeks, then judge honestly. If it saved real time, expand to the next task on your list. If it did not, you have learned something specific and cheap, rather than something vague and expensive.
The assumption is the risk
The businesses seeing revenue gains from AI are not more technical than yours, and their work is not more suited to it. They simply looked at their repetitive tasks and matched a tool to one of them, while the majority decided, without mapping anything, that there was nothing to look at.
“AI doesn’t apply to my business” is not a conclusion. It is a hypothesis, and it has never been cheaper to test.
Flux Dynamics helps UK SMEs find and build the AI use cases hiding in their day-to-day work, with a fractional CTO who builds rather than just advises. Start a project if you want a clear-eyed look at where AI actually fits your business.