22nd September 2026
Artificial intelligence is already being used by millions of people and thousands of businesses, but there is a surprisingly basic question that statisticians cannot yet answer properly.
How much is artificial intelligence actually contributing to the UK economy?.
The Office for National Statistics has now started work on a new way of measuring it.
At first glance this may sound like another statistical exercise of interest mainly to economists. In reality, it could become increasingly important because AI is beginning to change the way businesses operate, how people work and where companies invest.
The problem is that the traditional system used to measure the economy was designed long before ChatGPT, AI data centres and computer systems capable of generating text, images, software and other material became widespread.
The ONS says its existing framework does not explicitly identify AI, although areas such as software and research and development are already included in national accounts. The newer international System of National Accounts 2025 now gives much more explicit recognition to AI, data and cloud computing.
The ONS is therefore developing what it calls an AI thematic account.
What is an AI thematic account?
The phrase sounds complicated, but the basic idea is fairly simple. Think of the normal national accounts as a huge set of filing cabinets containing information about the economy. AI is already inside many of those cabinets, but it is mixed together with lots of activity that has nothing to do with AI.
The ONS wants to identify the AI-related parts and bring them together.
It will not replace the normal GDP figures. Instead, it will provide another way of looking at the economy and show how much activity can reasonably be attributed to AI.
The ONS says the work will initially concentrate on AI infrastructure, research and development and business investment before gradually expanding to downstream services using AI.
It is much bigger than ChatGPT
This is perhaps the most important point for ordinary readers.
When people hear AI, they often think about ChatGPT or similar programmes. The ONS's definition is much broader.
Its proposed measurement includes AI data centres and cloud computing, specialist computer chips, AI hardware, autonomous robots, vehicles and drones, software development, AI training and applications, cybersecurity and consultancy.
It also wants to capture areas where AI is being applied, including medical diagnosis, education, recruitment, engineering and other professional services.
So the economic impact of AI could eventually stretch from a giant data centre consuming enormous amounts of electricity to a small business using AI to deal with customer enquiries.
That makes measuring it particularly difficult.
There is another complication
AI can generate economic activity without all of that activity necessarily becoming British economic output.
Suppose a Scottish business pays for an AI service operated by an American company.
The Scottish business may become more productive. It may employ fewer hours doing a particular task, produce more work or offer a new service.
But the underlying AI technology may be owned outside Britain.
The ONS points out that major foundation-model providers, including OpenAI, Google and Anthropic, are US companies. Payments for access to their technology can therefore be treated differently from an AI asset developed and owned by a British company.
There is a similar problem with hardware.
Many of the specialised chips needed for AI computing are manufactured outside the UK and imported.
Britain may therefore gain substantial economic benefits from using AI without capturing the full value of the technology itself.
Britain is already adopting AI
This new statistical work comes as the ONS's business surveys show AI use increasing rapidly.
Its latest analysis found that the proportion of businesses with 10 or more employees reporting AI use had increased from around 12% in late 2023 to around 35% in 2026.
Adoption is much higher among larger companies, while there are substantial differences between industries. Around 58% of businesses in information and communication reported using AI, compared with about 13% in construction.
But even those figures need some caution.
The ONS points out that saying a company "uses AI" does not tell us how important AI is to its business. One company might use it occasionally to draft emails while another might have AI deeply embedded in its production systems.
The statistical challenge is therefore not simply how many businesses use AI, but how much economic activity depends on it.
And this could matter for productivity
Britain has struggled with weak productivity growth for many years.
If AI enables workers and businesses to produce more with the same amount of time and resources, it could eventually make a measurable difference to productivity and economic growth.
But that benefit may take time to appear in the statistics.
Businesses first have to invest in technology, train staff and reorganise how work is done. The gains may only become obvious once those changes have taken place.
The ONS itself says AI has the potential to change production processes, investment, labour demand, business organisation and productivity.
There is a Caithness connection
This may seem like a story about Silicon Valley rather than Caithness, but the implications are becoming increasingly local.
AI requires computing power. Computing power requires data centres. Data centres require large amounts of electricity, reliable grid connections and suitable land.
That brings the story back to the rapid expansion of Scotland's electricity generation and transmission infrastructure.
For areas such as Caithness, which are already seeing major renewable-energy and grid developments, the growth of AI could eventually create another source of demand for electricity and digital infrastructure.
It could also create opportunities for engineering, construction, telecommunications, maintenance and specialist services.
The question is whether those economic benefits will be captured locally or whether Caithness simply becomes another place producing electricity for businesses elsewhere.
The statistics are only just catching up
The ONS is not yet presenting a final figure for the value of AI to the British economy.
Instead, it is building the machinery needed to calculate it properly.
That may sound rather dry, but it is actually an important development.
Technological change can move much faster than economic statistics.
The ONS has provisionally set a timetable for developing its AI measurement, with further work on classifications during 2026 and a wider AI development plan expected by March 2027.
In a few years we may be able to look at the national accounts and see a much clearer picture of how much of Britain's economic activity is being generated, supported or transformed by artificial intelligence.
For now, perhaps the most interesting conclusion is that AI is already becoming part of the economy faster than we have been able to measure it.
The next question is whether that technological revolution eventually produces what Britain has been waiting for in another area for many years: stronger productivity growth and a larger economy without simply requiring more people and more hours of work.
Read the full ONS report HERE