# The small firms getting results from AI are using it on the boring jobs

> New UK research says the firms getting most from AI use it deeply on process-heavy work, not just chatbots. What that means for stock, orders and invoices, from Dan Cluderay.

Published: 2026-10-21
Author: Dan Cluderay

The firms getting real results from AI use it deeply, on process-heavy work. Pick one boring job, decide what the AI can see and change, and keep a person on anything that touches money or stock.

Two pieces of research landed on 19 October and they don't agree on much. The Social Market Foundation surveyed 1,015 small and medium-sized business leaders, and almost four in five say their business already uses AI tools. Enterprise Nation's figure is that only 21% of UK small businesses use it regularly. NatWest put it at 44% last month, which [I wrote about here](/insights/uk-small-firms-ai-adoption/).

The numbers differ because the questions differ. "Have you used ChatGPT" and "does AI do regular work in your business" are not the same thing. So I wouldn't get hung up on the headline percentage. The more useful findings are about who gets results.

## Depth beats dabbling

The SMF found that advanced users report significantly better results than basic users. It also found specialist tools did slightly better, and more consistently, than general chatbots. Revenue uplift depended more on what AI is used for, and how deeply, than on simply having it.

That fits with running a warehouse. Process-heavy work is where the repeats are. Checking stock against open orders. Spotting a PO that hasn't been booked in. Chasing an invoice that should have gone out when the order shipped. None of it is clever. All of it eats hours, every day.

A chatbot can help you write about those jobs. It can't do them unless it can see your systems.

## The worries are the right worries

The barriers people gave the SMF were data security (33%), quality or reliability (31%) and skills (26%). Enterprise Nation found 46% of small business owners don't trust AI tools enough to act.

I don't think that's a problem to talk people out of. If an AI is going to touch stock or invoices, you should be careful. The fix is to be specific:

- **One job at a time.** Start with something that repeats daily and is easy to check, like flagging orders that can't be picked.
- **Write down what it can see and change.** I set out how in [Before your AI touches stock or invoices, write down what it can touch](/insights/write-down-what-your-ai-can-touch/).
- **Keep a person on anything involving money or stock** until it has a track record.

## Where the data has to come from

Deep use only works if your systems are joined up. If stock lives in one place, orders in another and invoices in a third, the AI has the same problem your staff do. That's why most of our work starts with the plumbing, like [joining Shopify to the warehouse](/integrations/shopify-wms/), before anything clever goes on top.

It's also why we built [Bud](/bud/) to take action inside a business's own systems, with clear limits, rather than sit to one side as another chat window. More on that difference in [AI that takes action](/insights/ai-that-takes-action/).

## Start small, start this month

The SMF's modelling suggests wider adoption could be worth £38bn to £47bn in revenue to small and medium-sized businesses in Great Britain. Whatever the real figure, it won't come from more people asking chatbots for email drafts. It'll come from firms picking one boring job and handing it over properly.

If you want help picking that job, [get in touch](/contact/).

## Sources

- Social Market Foundation, Wider AI adoption could add £43bn to small business revenues, 19 Oct 2026: smf.co.uk/wider-ai-adoption-could-add-43bn-to-small-business-revenues-new-modelling-finds/
- Elite Business, Enterprise Nation and Visa Foundation launch $2 million initiative to close small business AI gap, 19 Oct 2026 (Enterprise Nation Tech Hub report figures)
- NatWest Group, AI Adoption Report press release, 25 Sep 2026