A third of firms now use AI, mostly for admin. The next win is in sales

By Jason Sibley

Two recent surveys seem to disagree. One says about a third of UK businesses use AI. The other says nearly four in five. They count different things, so both can be right.

What the government survey found

The GOV.UK AI adoption survey report 2026 executive summary says one-third of UK businesses were using at least one AI technology in 2026. That's 32%, against 16% in 2025, double the figure a year earlier. Larger firms were ahead at 52%.

Among businesses using or planning to use AI, the most common uses were:

• administration, at 81%
• marketing, at 71%
• IT functions, at 66%

The main driver was efficiency and productivity, at 71%. Limited AI skills, expertise or knowledge affected 67% of businesses that had adopted AI, and 78% of those planning to.

Most firms are starting with the back office. Many say they lack the know-how to go further.

Where SMF says the value is

The Social Market Foundation's Byte-sized growth, published on 19 October 2026, surveyed 1,015 UK SME leaders. It found that 79% said their business uses AI in some capacity. By size, that was 60% of micro firms, 86% of small firms and 92% of medium firms.

The gap between 79% and 32% needs care. I can't fully explain it, but the surveys measure differently. GOV.UK covers UK businesses of all sizes. SMF asked SME leaders, whose answers are self-reported. "Uses AI in some capacity" is also a looser bar than "using at least one AI technology". I wouldn't blend or average the numbers. Treat them as two views, and check each report's wording.

SMF says 87% of senior leaders reported a positive impact. The report gives 88% for time savings and 79% for quality; check it for who those percentages cover. It also says 55% reported a revenue increase. These are leaders' reports, rather than results SMF measured.

Advanced users reported significantly better impacts than basic users. Process-heavy departments such as HR, sales, supply chain and IT appear to be the highest-value applications. SMF also found AI adoption was 2.7 times more likely to leave employment unchanged than to increase it.

Why sales is the next step

This section is my opinion, not a survey finding.

Admin is the easy place to start. It's internal, low-stakes, and nobody outside the business sees a mistake. I'd guess that explains much of why it tops the GOV.UK list. But a tidier inbox rarely changes the bank balance by itself.

Sales is where small firms quietly lose money. The quote goes out and nobody chases it. The customer who bought twice last year hasn't heard from you since. The happy client never gets asked who else might need you. These jobs repeat and follow a clear pattern, which suits AI drafting. SMF's finding that sales is a high-value area fits what I see.

This approach also fits the skills problem: one small task, done carefully, needs no big project. For a method, see one AI workflow for 30 days.

Three sales jobs to try this month

Use each prompt in whichever AI tool you already use. Replace everything in square brackets with your own details.

Follow-up after quotes

Most quotes don't get a reply on the first try, and most of us stop chasing too soon. Our piece on sales follow-up that survives the inbox covers the habit. For the quote itself, see AI quotes and proposals. If you track deals in a CRM, the tiny pipeline guide for HubSpot shows a simple way to do it.

I run a small UK business. A customer hasn't replied to a quote, and I'd like a polite follow-up email.

Write three short versions, in British English, with a warm, plain tone:

- a gentle nudge a few days after the quote
- a second follow-up that adds something useful
- a final friendly message that closes the conversation

Do not invent any facts, prices, deadlines, discounts or customer details. If you need something I haven't given you, write [GAP: what's missing] so I can fill it in.

Quote details: [paste here]

Notes on our conversations so far: [paste here]

Reactivating lapsed customers

Pick customers who bought from you before but have gone quiet. The guide to AI for customer emails has more on tone.

This isn't legal advice. If you're unsure who you're allowed to email, ask a professional adviser before sending anything.

I run a small UK business. I'd like a short, friendly email to a past customer who hasn't bought from us for a while.

Write one email of under 120 words, in British English, with no hard sell. Say we've noticed it's been a while, mention what they bought before, and offer one simple next step.

Do not invent any facts, offers, prices, dates or details about the customer. Where something is missing, write [GAP: what's missing] for me to fill in.

What they bought before: [paste here]

When they last bought: [add here]

What we can offer now, if anything: [add here]

Asking for referrals

The best moment is straight after a good result. Our piece on the referral ask after a win covers the timing.

I run a small UK business. A customer has just told me they're happy with our work, and I'd like to ask if they know anyone else who might need us.

Write a short, warm message in British English that thanks them, makes the ask easy to say yes or no to, and doesn't pressure them.

Do not invent any facts, quotes, results or details about the customer or the work. Mark any gaps as [GAP: what's missing] for me to fill in.

What we did for them: [add here]

What they said: [paste their words here]

Keep a person on it

Read everything before it goes out, and send it as yourself.

AI can write a lovely email with the wrong price, a deadline you never offered, or a "great to see you again" for someone who's never bought from you. Customers notice those slips, and sales is the worst place for them. A quick check before sending covers most of it:

• [ ] Names, prices and dates are correct
• [ ] Nothing is promised that you can't deliver
• [ ] Every [GAP] has been filled in or removed
• [ ] It sounds like something you'd actually say
• [ ] You're happy to be held to every line

The 55% revenue figure from SMF is what leaders reported. It doesn't prove AI caused their increase or will cause one for you. Try one job and judge it by what happens to your replies and bookings.

If you'd like company on this

At My AI Apprentice we run AI sessions for small businesses, including ones focused on sales follow-up like the jobs above. No pressure either way.