AI is the best thing that's ever happened to small business

Most of the AI conversations in the national press are about the wrong thing.

It’s about which lab is ahead. Whether the UK missed the boat because the big models are all built in San Francisco. Whether we need our own sovereign LLM. Is our energy cheap enough to run data centres.

None of that matters if you run a business. You are never going to train a model. You are going to use one. The country that builds them and the country that gets rich from using them do not have to be the same, and the businesses that get the most from using them are not going to be the ones with the biggest R&D budget.

The numbers nobody puts in the headlines

The UK data is genuinely encouraging if you know where to look.

The Office for National Statistics has tracked business AI use since 2023. Among firms with 10 or more employees, self-reported use has gone from about 12% to about 35% in under three years. The British Chambers of Commerce, working with the University of Essex, puts active adoption among SMEs at 54% in 2026, up from 35% in 2025, 25% in 2024 and 23% in 2023.

Adoption more or less doubles every 18 months, which is really positive.

The jobs panic hasn’t materialised either. 95% of AI-using SMEs told the BCC that AI had no impact on their workforce size over the past year. 86% said job roles were unchanged. The Bank of England’s Decision Maker Panel found the same thing across three years of data.

So the picture is: more businesses using it, roughly nobody firing people because of it, and three quarters of adopters reporting a productivity lift.

It’s even better for SMEs

Here is the part that should make you sit up.

MIT’s NANDA study looked at 300 public AI deployments and found that 95% of enterprise generative AI pilots delivered no measurable impact on profit and loss. Not 5% failure. 95%.

The reasons are dull and structural. Legacy systems that won’t talk to anything. Data sitting in four different places with no owner. Procurement cycles. Legal review. A steering committee. A change management programme. A pilot designed to impress a board rather than fix a job someone actually does.

The ONS number that says the most is this one: the average AI-adopting business has gone from using 1.4 AI tools to 1.6 since 2023. Three years of hype, 0.2 of a tool.

A UK benchmark report put the value going unrealised at around £78 billion. Only 31% of businesses report a positive return on their AI spend, while 85% to 91% keep increasing the budget.

None of that matters for SMEs. You do not have a steering committee. If you decide on Tuesday that your quotes are going to be written differently, they are written differently on Tuesday.

What one person can now do

Three years ago, a sole trader who wanted to compete properly needed to hire, outsource or sacrifice every waking hour. A copywriter for the website. A bookkeeper for the receipts. Someone to answer the phone. A designer. A researcher.

Now, one person with a decent setup handles all of it before lunch.

Some real examples of what this looks like in practice:

• A tradesman who spends two evenings a week writing quotes now dictates the job into their phone on the drive home and gets a formatted quote in their own tone of voice, ready to check and send.
• A small consultancy that limited public tender submissions because the paperwork took three weeks now bids on four times as many, because the first draft takes 40 minutes.
• A brand owner who never had time for a newsletter now writes one a week from voice notes taken during the day.

The St. Louis Fed measured the time saving at around 2.2 hours a week for the average worker using generative AI. That is a national average across everyone, including people using it once a fortnight to rewrite an email. For someone who sets it up properly around the jobs they actually do, the number is much bigger. In my experience with clients, the first properly built workflow usually saves between four and eight hours a week on its own.

For a business of one to ten people, eight hours a week is close to a fifth of a hire, for the cost of a phone contract a month.

So AI can do it, can you?

Small businesses have the structural advantage but they are not currently using it.

Cambridge’s Bennett School analysis of ONS firm-level data shows large firms (250+ employees) nearly doubled adoption to 44% between 2023 and 2025. Small firms, under 50 employees, reached 26%. Micro businesses are at roughly 14%.

So the firms with the red tape are moving faster than the firms without it. The gap is widening.

The reason isn’t cost. The tools cost £20 a month. What’s missing is somebody sitting down with a business owner for two hours and showing them how it fits around what they already do.

Being AI native

The businesses that pull ahead over the next five years won’t be AI companies. They’ll be the ones that happen to run every part of their operation with AI in it.

That means it shows up in the quoting, the marketing, the invoicing, the customer follow-up, the tender writing, the hiring and the reporting. Not one chatbot bolted onto the website.

The dotcom comparison gets made a lot and it’s roughly right, with one difference: the businesses that won the web era needed a developer, a budget and 18 months. This time the barrier is a Wednesday evening and someone to show you where to start.

Where to actually start

If you run a small business in the South West and you’ve been putting this off, do these three things this week.

1. Write down the job you hate most. The one that eats an evening: quotes, invoices, chasing, the monthly report. One job, the one that annoys you.
2. Do that job once with an AI tool sat next to you. Claude, ChatGPT or Gemini, £20 a month, no setup required. Talk to it like you’d brief a new starter. Give it three examples of how you normally do it.
3. Save what worked. The prompt that got it right is now a tool you own. Use it again next week. That is the whole thing.

Bonus: create a skill, custom GPT or Gem so you don’t have to type the prompt again.

Bonus bonus: schedule the task, so you don’t even have to type again.

Most people stop after step two because the first attempt sounds like a robot wrote it. It does, because you gave it no examples of your voice. Give it three, and it stops.

Three years in, the technology has gone from a novelty to something that genuinely changes what a small team can carry. It is going to keep accelerating. The businesses that adopt it early and properly are going to be difficult to compete with.

If you want a hand working out where it fits in your business, that’s what I do. First conversation is free and takes half an hour.

---

Sources: ONS Business Insights and Conditions Survey (2023–2026); British Chambers of Commerce and University of Essex MiSoC, Future of Work: AI in the Workplace (March 2026); MIT NANDA, The GenAI Divide (2025); Bennett School of Public Policy, Cambridge (April 2026); Federation of Small Businesses; Federal Reserve Bank of St. Louis (2025).