Most conversations about AI adoption lean on a single national number, and that number tends to make business owners feel behind. New data from the Office for National Statistics tells a more useful story, because it breaks that number apart by sector, and the picture underneath is very different from the headline.

The ONS Business Insights and Conditions Survey, published in July and covered by Business Cornwall in August, found that 35% of UK businesses with ten or more employees now use at least one AI technology. Look past that average and the spread is stark. Information and communication leads at 58%. Education sits at 51%. Professional, scientific and technical services follow at 46%. At the other end of the same table: wholesale and retail on 21%, transport and storage on 19%, and construction on just 13%.

For a business in retail, logistics, construction or the trades, that changes the question worth asking. It isn't "why haven't we caught up with everyone else," because there is no single "everyone else." It's "where does our own sector actually stand," and for most of these industries, the honest answer is that it has barely started.

Why the low number is the interesting part

It's tempting to read a 13% adoption rate in construction as evidence that AI simply doesn't apply to that kind of work. That's rarely the real reason. Low sector-wide adoption usually reflects fewer businesses that size having sat down and worked out where AI would actually help them, not an absence of anywhere it could. Estimating, scheduling, supplier communication, site reporting and quoting all generate the kind of repetitive written and numerical work that AI tools handle well. The gap isn't capability. It's that almost nobody in the sector has looked yet.

That's also what makes a low sector average a genuine opportunity rather than a reason to keep waiting. When close to nine in ten businesses in your industry haven't adopted any AI technology, meaningful movement doesn't require a transformation programme. It requires being one of the businesses that looked properly, before the sector average catches up and a head start becomes a standard expectation instead of an edge.

The comfort of "our industry doesn't do this" runs out quietly

The businesses that end up furthest behind aren't usually the ones that tried something and it didn't work. They're the ones who assumed their industry was exempt for long enough that a handful of competitors quietly got ahead, then found the gap harder to close than it would have been a year earlier. Sector-wide adoption figures rise gradually, then the businesses still waiting notice all at once that "nobody in our trade does this" stopped being true some time ago.

Where to actually start

None of this calls for guesswork about which tools to buy or a leap straight to implementation. It calls for an honest look at where AI would genuinely take friction out of the way a specific business already works, which parts of the operation are ready for it, and which aren't, and what the realistic first step looks like. That's the structure an AI Accelerator Diagnostic works through: a half-day assessment across six dimensions, from vision and opportunities through to governance and value, ending with a written roadmap and one clear action to take in the coming week.

A low sector adoption rate isn't a signal to relax. It's a signal that the businesses who look now, while the rest of the sector is still working out whether this applies to them, get to set the pace rather than chase it.

Go wisely.

ONS Business Insights and Conditions Survey, published 20 July 2026