Web & Roll - Logo
AI Automation Sheffield

Half of UK small businesses now use AI. One in five feels confident with it.

47% of small business owners now use AI, up from 22% last year. But only 19% feel confident with it and a third use it for admin alone. Here is where Sheffield firms are leaving value on the table.

Categories
AI Solutions
Tom BarberTom Barber
Published
Reading time4 min

AI Solutions

Simply Business surveyed 1,842 UK small business owners over the summer and published the findings in September. 47% of owners now use AI tools, up from 22% a year ago. Adoption has more than doubled in twelve months.

The more telling numbers are further down the report. Only 19% of owners call themselves very confident using AI. A third use it for routine admin and nothing else. Among those not using it at all, 39% say they cannot see a clear use case.

So adoption has stopped being the question. Depth is the question.

What everyone is actually using it for

Content creation 63%. Problem solving 53%. Generating ideas 50%. Saving admin time 46%.

Every one of those is a job a public chatbot does on its own, with no setup and no connection to the business. Genuinely useful, and worth the subscription. But it is also exactly what your competitor down the road can have running by this afternoon for twenty pounds a month. It is a baseline, not an advantage.

Why the generic tools stop there

ChatGPT, Claude and Gemini are remarkable, and they have three limits that no amount of better prompting fixes. They do not know your products, your customers or the way you talk. They forget everything between sessions unless someone sets them up not to. And they cannot do anything in your other systems without integration work.

For a sole trader drafting emails, none of that matters. For a ten-person business with a pipeline and a CRM, those three limits are the whole gap between AI helping you write and AI doing the work.

Three patterns that earn their keep

Working with firms across Yorkshire, three use cases come up again and again:

  • A chatbot trained on your own documentation rather than the public web, answering product, pricing and process questions out of hours. It takes real load off the contact form and the phone.
  • Automations that move data between tools you already pay for. An enquiry landing on your site, going into your booking system, into your accounts, and into the right Slack channel, with nobody touching a keyboard.
  • Content drawn from your own projects, photos and the language your customers actually use, rather than from whatever the model guessed.

What bespoke actually means

It does not have to mean expensive or experimental. In practice:

  • Find the five tasks across your team that take the most time and repeat the same shape every week. Those are almost always the best candidates.
  • Decide between a managed platform like Make or n8n and bespoke development, based on how complex the job is and how sensitive the data is.
  • Connect it to your real sources, the CRM, the support inbox, the document library, instead of asking a general model to guess.
  • Keep a human checkpoint for the first month, then ease off once the output has earned trust.
  • Measure hours saved on named tasks and how often it gets things right, rather than productivity in the abstract. If you cannot see it in your reporting, you cannot tell whether it worked.

Where to start

That 39% who cannot see a clear use case are not wrong to be cautious. They are describing a mapping problem rather than a technology problem, and it is the easiest part to fix. A typical first project with us is two to four weeks of work aimed at one repetitive job, so you find out whether it pays before committing to anything bigger.

Where Web & Roll can help

We build practical, business-specific AI for Sheffield and Yorkshire firms: chatbots, automations and content systems grounded in your own data. It starts with a free chat to work out which jobs are worth automating and, just as importantly, which are not.

Sources