29%
of UK businesses with fewer than 10 staff used some form of AI in September 2026.
71%
of businesses asked, mostly ones not using AI, said they hadn’t found a use for it.
77%
of businesses using AI had seen no change in revenue yet.
14%
more customer problems solved per hour when support staff had AI help, and 34% more for newer staff.
Using It Isn’t the Same as Getting Paid Back
Most businesses that use AI use it to write: 85% of adopters in the government’s survey use it to generate text. That saves someone time. It rarely changes what a business earns.
In Denmark, researchers matched use of ChatGPT-style assistants with pay records. Two years after ChatGPT arrived, they found no meaningful effect on earnings or hours.
Where the Gains Show Up
Where gains were measured
- Customer support agents answering real queries, 14% more resolved per hour
- Consultants on tasks AI handles well: 12% more tasks, 25% faster, over 40% better quality
Where they weren’t
- The same consultants on a task outside what AI handles well were 19 percentage points less likely to get it right
- Workers using chat assistants in Denmark: no measurable change in pay or hours
The gain comes from the job, not the technology. AI pays where it is put to work on a specific, repeated job.
The ‘It Depends’ Test
Here is how we decide where AI belongs. Write the job down as one sentence.
Send a reminder seven days after an invoice is due.
The sentence finishes.
Plain automation. No AI needed.
Decide whether to chase or call, and what to say. It depends on who the customer is, what they’ve paid before and what their last email said.
The sentence stops at ‘it depends’.
That’s where AI helps.
Most of a job runs as plain rules. AI earns its place in the part that needs judgement, and a person approves what matters.
Three Questions Before You Pay for Anything With AI in It
- Which job does it do, start to finish?
- What number will move, and how will we both see it move?
- Where does my data go, and is it kept?
If the answer to the first is ‘it can do lots of things’, keep your money.
Sources
- ONS, Business insights and impact on the UK economy, 8 October 2026
- DSIT, AI adoption research, 28 January 2026
- Brynjolfsson, Li and Raymond, Generative AI at Work, Quarterly Journal of Economics, 2025
- Dell’Acqua and others, Navigating the Jagged Technological Frontier, Harvard Business School working paper 24-013, 2023
- Humlum and Vestergaard, Large Language Models, Small Labor Market Effects, NBER working paper 33777
Read next
- How We WorkPick one problem, agree what should change and how we’ll both know, and we fix it in an agreed number of days. Where AI helps, and where it doesn’t.
- What a missed call costs a small businessNobody publishes a reliable UK figure for missed calls. Here is how to count yours in a week, and work out what they cost you in a year.
Pick One Problem
Tell us which one. We’ll tell you whether we can fix it and how long it takes.