London businesses are more likely to use artificial intelligence than Scottish firms, according to figures from the Department for Science, Innovation and Technology. The data put AI adoption in London at 20%, compared with a UK average of 16%.
Scotland ranked lowest in the regional breakdown, with 84% of firms saying they were neither using AI nor planning to do so. The figures also pointed to a separate issue around readiness: Welsh adopters were the least likely in the UK to say they had the foundations needed to move AI beyond pilot schemes.
Colibri Digital, a consultancy working on AI projects in sectors including financial services, healthcare, energy and transport, said the gap reflected uneven access to resources rather than a lack of interest outside the capital. It said the bigger concern for businesses and policymakers was whether organisations had modern data systems, cloud infrastructure and operational controls in place to put AI into day-to-day use.
The data suggest adoption alone does not tell the full story. In Wales, only 34% of AI adopters said they felt ready to scale their systems, indicating that many companies may still be experimenting rather than embedding the technology in core operations.
Productivity gap
The regional divide matters because businesses already using AI are reporting measurable gains. Across the UK, 56% of firms that have adopted AI said they had seen productivity improvements, while government estimates put the potential annual economic benefit from AI innovation at £47 billion over the next decade.
Those gains could become concentrated in places that already have an economic advantage. London is already 28.5% more productive than the UK average in output per hour worked, and stronger AI uptake in the capital risks reinforcing that lead if other regions struggle to move from trials to wider deployment.
The issue reaches beyond the technology sector. Some of the largest and most economically important uses of AI in Britain are in industries spread across the country, including rail, healthcare, manufacturing and finance.
Network Rail is using AI-based predictive maintenance across track infrastructure, where faults account for about 341 days of delay a year. In healthcare, 99 of 107 English stroke units have access to AI decision support. UK manufacturers are also reported to lead Europe, with 53% adoption of AI on the factory floor, while three-quarters of UK financial firms are using AI.
If companies in those sectors cannot deploy systems at scale, the economic benefits forecast from AI are less likely to reach regional industrial centres, where much of the country’s infrastructure, workforce and production base are located.
Resource divide
Marvin Gillibrand, Head of Applied AI at Colibri Digital, said the pattern in the government data matched what the company had seen in client work over the past 18 months.
He said: “The firms scaling AI successfully are the ones that have done the unglamorous work upfront. They have modernised their data platforms, invested in cloud infrastructure and built the operational controls needed to deploy AI safely into production. Critically, they have also developed the experience and organisational confidence to move subsequent AI initiatives much faster.
“London’s lead in the DSIT data probably reflects a concentration of resources more than a concentration of ambition. Larger enterprises, deeper technology budgets, stronger access to AI talent and earlier cloud adoption all make it easier to operationalise AI at scale.
“For organisations outside London, the challenge now is not proving the value of AI in principle. It is building the operational maturity needed to move reliably from experimentation into production.
“The good news is that AI capability is no longer confined to London. Cloud infrastructure, open-source models and managed AI platforms have lowered the barrier to entry significantly over the past few years. Some northern organisations are in a strong position because they tend to be more operationally focused and closer to real industrial use cases in sectors such as energy, transport and manufacturing.
“The firms moving quickest are often the ones taking a pragmatic approach: modernising data foundations, targeting a handful of high-value use cases and building confidence incrementally rather than trying to transform everything at once.”
The findings add to a wider policy question over whether Britain’s AI strategy will deepen or narrow regional economic disparities. While interest in AI appears to be spread broadly across sectors and geographies, the ability to support real-world deployment remains uneven, with infrastructure, skills and technology budgets still concentrated most heavily in and around the capital.
For businesses outside London, the data suggest the challenge is no longer simply deciding whether to adopt AI. It is whether they can build the systems and controls needed to make that adoption count in production environments where productivity gains are realised.
