The headline arrives; the high street is busy
Somewhere on Grantham's high street today, a café owner is managing rotas on a whiteboard, reconciling card receipts on a phone, and half-reading a newsletter about how artificial intelligence will transform small businesses. The pitch is familiar by now: save hours every week, automate the routine, compete with the big players. What the newsletter does not mention is who it was written for — and it was not written for her.
That gap is what this piece is about. Not whether AI is impressive (it is, in places), and not whether small businesses should ignore it (they probably shouldn't). The question is more specific: when the tools land in a market town like Grantham — independent shops, sole traders, hospitality businesses, the kind of high street that runs on personal relationships and tight margins — do they actually fit? Grantham is a useful place to ask, precisely because it is not London, not a university city, and not a tech cluster. It is, instead, a fairly accurate picture of where most British businesses actually operate.
Most UK small businesses haven't adopted AI — and most aren't planning to
The national numbers are more striking than the AI headlines suggest. According to DSIT's 2025 survey of 3,500 businesses, only one in six UK businesses uses any AI technology at all — and even that figure comes with caveats: adoption was described as patchy, mostly small-scale, and confined to experiments rather than embedded in how businesses actually run.
YouGov's concurrent survey of SME decision-makers is sharper still. Just 5% use AI extensively. Forty-four percent don't use it at all. And 80% say they are satisfied with the software they already have. Perhaps the most telling figure: 54% do not believe AI will replace traditional business tools within three years — the very timeline AI vendors typically cite to create urgency.
That is not a portrait of businesses straining to get on board. It is a portrait of businesses that have looked at the tools on offer and, on balance, decided the case has not yet been made.
Shadow AI complicates the picture slightly. DSIT's methodology did not capture informal use — the employee who drafts emails in ChatGPT, the sole trader who occasionally pastes a problem into a free AI chat tool. Factoring that in, true usage is probably higher than 16%. But individual, ad hoc use of a consumer tool is not the same as productive, managed AI adoption. It does not save five hours a week; it saves five minutes, unpredictably, when someone remembers to try it.
The software wasn't designed for this kind of business
The barriers that SMEs name in surveys are not what the 'just try it' narrative tends to assume. Reliability and accuracy concerns top the list — cited by 65% of non-adopters — followed by data security and GDPR (51%), regulatory complexity (33%), and the cost of switching from existing software (30%). These are not expressions of technophobia. They are rational assessments of tools that were not, in the main, built for businesses of this size.
The data problem is structural. AI tools work best when the information they process is clean and consistently organised. A sole trader's operational data rarely is: invoices spread across PDFs and paper, customer preferences held in memory, verbal supplier agreements recorded nowhere. Feeding fragmented, unstructured information into an AI tool does not produce reliable outputs — it produces confident-sounding mistakes. Which is precisely what that 65% reliability concern reflects.
Then there is the integration barrier. Using AI productively — rather than simply generating a draft that still needs heavy editing — means connecting the tool to actual business data: accounting software, booking systems, customer records. Those connections require API integrations. API integrations require IT skills or IT support. A sole trader working alone has neither, and a training course does not change that.
The deeper issue is architectural. Most commercial AI products were designed around enterprise assumptions: a compliance team to manage GDPR risk, an IT department to handle integration, and enough structured data to make the system useful. Strip those out — as you must with a two-person café or a market-stall holder — and what remains is a product that presupposes an organisational structure that does not exist. That is a design mismatch, not a skills gap. No amount of upskilling provision resolves an architecture problem.
Smaller businesses in smaller places fall furthest behind
Size and geography compound each other rather than sitting as separate disadvantages. Enterprise Nation's survey of 1,000 SME decision-makers found that 37% of sole traders and 25% of micro businesses do not use AI at all, against 19% of SMEs overall. Those are precisely the business types that populate a market town high street: the independent retailer, the two-person café, the market stall trading on the same pitch for decades.
The gap does not close once a business does adopt. Among micro firms that use AI, only 37% automate tasks through it; among medium-sized firms, that figure is 54%. Fewer than one in five micro businesses deploy AI agents at all, compared with roughly one in three medium firms. Crossing the adoption threshold is only the first hurdle; depth of use falls away sharply at the smaller end.
Geography reinforces this. London records 93% AI usage across businesses. For firms in non-metropolitan regions comparable to the East Midlands rural fringe — Yorkshire and Humber, the South West, Scotland — non-adoption runs at 24–28%. No Grantham-specific survey exists; that is an extrapolation from comparable regional data, not a local count.
Age adds a further layer. Forty percent of SME decision-makers aged 55 or over do not use AI, against fewer than 10% of those aged 18 to 34. Older business-owner profiles are more common in market towns — a factor worth recording without inflating into a cultural diagnosis.
The shadow AI caveat applies here too. Because informal tool use is largely uncounted, the real figure for a town like Grantham is genuinely uncertain — and that uncertainty does not default to optimism. Unmanaged, occasional use of a consumer tool is not productive adoption; it simply makes the gap harder to measure.
What Grantham's local infrastructure can and cannot offer
Grantham does have genuine local assets. Grantham College became a founding node in the Lincolnshire Institute of Technology (LIoT) — a DfE-backed consortium led by the University of Lincoln — and now offers Level 4 Digital Technologies qualifications covering AI, cyber security, and data analytics. That is a real provision, not a marginal one.
The scope, however, is specific. The LIoT's employer partners are Bakkavor, Siemens Industrial Turbomachinery, Olympus Automation, and Autocraft — the manufacturing and industrial workforce for which the consortium was designed. Its qualifications are calibrated to factory-floor and engineering contexts. An independent café operator, a market-stall trader, or a sole-trader retailer is not the intended beneficiary. This is not a criticism of what the LIoT is doing; it is a description of what it is not doing, and that distinction matters when assessing what upskilling provision is realistically available to the high street.
The £5.56m Future High Streets Fund and the South Kesteven Town Centre Action Plan are likewise genuine commitments — directed at public realm improvements and economic regeneration, not digital capability. They address footfall and environment, not whether a sole trader can navigate GDPR in a cloud-based AI tool.
Connectivity adds a structural constraint. Roughly 12% of rural premises nationally lack access to basic 10 Mbps broadband, and Grantham's semi-rural hinterland shares the mobile 'not-spot' problems common across the East Midlands rural fringe. Cloud-based AI tools depend on reliable connections; intermittent coverage is not a minor inconvenience for a business that operates from a market pitch or a rural unit.
Taken together, these gaps point toward a plausible risk that the local evidence already names: a two-tier economy in which AI dividends accumulate with technically equipped external suppliers and industrial employers, while independent high-street businesses face compounding digital exclusion as commerce and public services shift to digital-first models. That outcome is not inevitable, but the local infrastructure as currently configured does little to prevent it.
What realistic AI adoption actually looks like from here
So, to return to the café owner with the whiteboard: some tools are probably already within reach. Drafting a reply to a TripAdvisor review using ChatGPT, generating a week's worth of social media captions, or prompting a language model for a supplier email — none of that requires API integrations, clean structured data, or an IT department. This is where shadow AI is already happening, quietly, on individual phones and laptops, and it is a reasonable place to start.
Beyond that narrow band, the honest answer to the article's opening question is that the gap is real and structural. It is not primarily a training-course problem or an awareness problem. The tools were built for organisations with compliance teams, data hygiene, and IT support — not for a sole trader whose business records live across three apps, a receipt drawer, and a notebook.
What would genuinely help is different from what is currently on offer: software designed to work with messy, fragmented data rather than against it; digital support embedded in local business networks rather than delivered only through college qualifications aimed at the manufacturing sector; and vendors willing to set realistic expectations about what small-scale, low-infrastructure adoption can actually deliver.
The argument is not that AI has nothing to offer market towns. It is that the current tools were not designed with them in mind — and until that changes, the productivity gains will keep landing somewhere else.