NatWest survey finds most UK business AI use stays shallow

NatWest Group says 44% of the UK small and mid-sized businesses it surveyed are now using artificial intelligence. That is the headline figure in the bank's first AI Adoption Report, published on 25 September 2026. The more revealing number sits further down the page: of those current users, only 6% have reached what NatWest calls the "transforming" stage, where AI has changed how the business actually operates. Another 29% of users are still at the earliest stages of awareness-building or limited piloting.

The findings matter to UK SMEs and mid-market businesses, because they describe a familiar pattern in new numbers: access to AI tools is spreading fast, but turning that access into changed operations is happening much more slowly. AI use, system integration and formal governance are all higher among larger firms, official data shows. Independent UK data broadly backs that shape of the story, even though it does not confirm NatWest's specific figures.

This is a bank-commissioned survey, not an official statistic, and it comes with a commercial backdrop: NatWest sells business banking, and promotes a business-support programme, a partnership with Google Cloud and its own AI strategy alongside the report. That does not make the findings wrong, but it does mean the numbers deserve the same scrutiny a reader would apply to any research published by a company with a stake in the answer.

What NatWest actually measured

NatWest's business findings come from a survey of 1,400 UK SMEs and mid-market businesses, run alongside interviews with large corporate leaders, a survey of more than 2,400 NatWest Retail customers, and a benchmark against a nationally representative sample of 1,800 UK consumers. The press release and the report page were both published on 25 September 2026.

Two things are missing from the public material. First, NatWest does not state the business survey's fieldwork dates, research agency, sampling frame, response rate or weighting on the pages reviewed for this article, and it does not say whether respondents were existing NatWest customers. Second, while NatWest calls its 1,800-person consumer sample "nationally representative", it does not make the same claim for the 1,400-business sample. Until NatWest publishes a fuller methodology, the 44% figure should be read as a result from NatWest's own sample, not as an estimate of AI use across all UK businesses.

There is also an internal inconsistency worth flagging. NatWest's report page and its press-release headline treat the 6% "transforming" figure as a share of current AI users. The body text of the press release instead describes it as 6% of businesses overall. Because 56% of NatWest's sample are not yet AI users at all, that is a meaningfully different number, and the packet reviewed for this article could not resolve which framing NatWest intends as definitive.

Widening, not yet deepening

NatWest's own breakdown of adoption stages shows the shape of the problem: 44% of surveyed businesses use AI, a further 41% expect to adopt it within five years, but among today's users, 29% are still at the earliest awareness or limited-pilot stage and only 6% have reached transformational use.

Official UK statistics describe a similar pattern, even though the exact percentages differ because every survey defines "using AI" differently. The Office for National Statistics, in analysis published on 20 July 2026, found that AI use among UK businesses with at least 10 employees rose from around 12% in late 2023 to around 35%, while the average number of AI technologies each adopting business used rose only from about 1.4 to about 1.6 — evidence, in the ONS's own words, that adoption remains "relatively shallow". The Department for Science, Innovation and Technology's UK Business Data Survey 2026, fielded between October 2025 and January 2026, found that 41% of businesses handling digitised data used AI, but only 21% of those users had integrated AI into their existing business systems.

None of these figures are interchangeable. They cover different populations, different fieldwork periods and different definitions of adoption. But taken together, they support the broad conclusion that trying AI is becoming common while embedding it into how a business runs is not.

The size divide

NatWest reported AI use of 67% among businesses with more than 100 employees, against 36% among smaller respondents. It also said its statistical model attributed 39% of the variation in AI use to headcount, 24% to turnover and 8% to geography — though the pages reviewed for this article do not explain the model behind those figures, so they should be treated as a reported result rather than an independently verifiable calculation.

Official data points the same way using different size bands. ONS found 28% AI use among businesses with zero to nine employees, against 49% among those with 250 or more. The UK Business Data Survey 2026 found system integration ranging from 18% among sole traders and 27% among microbusinesses up to 57% among large businesses, and it found formal written AI policies in just 3% of sole traders and 8% of microbusinesses, compared with 56% of large businesses.

The Office for National Statistics' 2023 work on management practices found that, among businesses citing barriers to AI adoption, 39% pointed to difficulty identifying relevant use cases, 21% cited cost, and 16% cited a lack of AI expertise and skills. Both NatWest's and the official data associate larger business size with higher AI use, though neither source establishes why that association holds.

Human accountability, not just automation

Among the more than 2,400 NatWest Retail customers surveyed, 81% said the ability to reach a real person when needed was the most important trust factor in AI-enabled financial services. 77% identified complete service reliability, human oversight and transparency in how their data is handled as critical, and 75% regarded independent regulatory oversight as important.

These figures describe NatWest's own retail customers, not the separate 1,800-person nationally representative consumer benchmark, and the material reviewed does not show whether the two groups answered differently. Even so, the results show that NatWest's customers place high value on human access, reliability, oversight and transparency, and on independent regulatory oversight, when using AI-enabled financial services. The survey itself does not establish what rights or protections apply in any given case; a reader with questions about a specific product or provider should check the current position directly with the relevant regulator, such as the Financial Conduct Authority or the Information Commissioner's Office.

Do deeper adopters really gain more?

NatWest reported that businesses at its earliest adoption stage typically claimed working-time savings of 1% to 10%, while those at the "transforming" stage reported average savings of 61% to 70%. It also reported increased revenue among 95% of transforming-stage businesses, against fewer than 29% at the earliest stage.

These are self-reported, respondent-claimed outcomes, not independently measured productivity figures. NatWest has not published the number of respondents behind the transforming-stage subgroup, a confidence interval, or any attempt to separate the effect of AI from the effect of already being a larger, better-resourced or better-managed business. ONS's own work on management practices treats technology adoption as associated with — not proven to cause — stronger turnover per worker and better management quality. It is entirely possible that firms which are already more capable are simply more likely both to adopt AI deeply and to report strong results, rather than AI causing those results on its own.

How much weight to give a bank's own survey

NatWest's report is useful as a snapshot of its sample of 1,400 UK SMEs and mid-market businesses, and its broad direction — widening access, shallow depth, a real gap between larger and smaller firms — lines up with official ONS and DSIT evidence gathered independently. But it is commissioned research from a bank that sells business banking and AI-related services, published without the fieldwork dates, sampling detail, weighting or full adoption-stage definitions that would let a reader check the numbers independently. Readers should treat the 44% and 6% figures as directionally consistent with the wider UK picture, not as precise, peer-reviewed estimates of national AI adoption.

What to watch next

No full methodology or technical appendix appeared on the NatWest public pages reviewed for this article; publication of one would allow independent scrutiny of the sample, weighting and adoption-stage definitions behind the 44% and 6% figures. Future updates to the UK Business Data Survey or ONS's AI in UK businesses analysis, if published, could show whether system integration and formal AI governance move beyond the low levels — 21% integration, 5% formal written policy — recorded for 2025–2026. Businesses considering AI deployment that touches customer data or regulated financial decisions should check current guidance from the Financial Conduct Authority and the Information Commissioner's Office rather than relying on adoption statistics alone.