The Office for National Statistics has begun building a new framework intended to measure artificial intelligence as an identifiable part of the UK economy, addressing a growing gap between rapid adoption and the statistics used to track its economic effects.
The ONS published its methodology for an AI thematic account on Monday, setting out how it intends to identify AI-related economic activity within the national accounts.
The work does not yet provide an official estimate for AI’s contribution to GDP. Instead, it establishes the definitions and accounting structure needed to measure production, investment, use, labour effects, and other economic activity more consistently as better data become available.
Thematic accounts reorganise economic information around subjects that are difficult to isolate within conventional industrial classifications. AI presents that problem because its development and use extend across software, cloud infrastructure, finance, manufacturing, professional services, healthcare, government, and other sectors.
A specialist model developer can be identified relatively easily as part of the AI economy. A manufacturer using machine-learning systems to improve production, or a financial-services company embedding AI into an existing workflow, is harder to separate from the surrounding economic activity.
National accounts can record the value of the software, equipment, labour, and services involved without necessarily identifying which share relates specifically to artificial intelligence. The ONS thematic account is intended to make more of that activity visible without replacing the existing national accounting framework.
The initiative comes as spending on AI increasingly extends beyond experimental software licences. Organisations are investing in data infrastructure, compute, model access, integration, governance, cybersecurity, workforce skills, and redesigned business processes.
Those costs can appear across different accounting categories even when they support a single AI programme. That makes comparisons between industries and organisations difficult, particularly when private-sector estimates use different definitions of what constitutes an AI company or AI investment.
A clearer statistical framework could also strengthen analysis of productivity. AI can alter the amount of labour required for an activity, improve the quality or speed of an output, or enable entirely new services. Those effects may appear elsewhere in economic statistics without being straightforwardly attributable to the technology.
The measurement challenge is becoming greater as AI becomes embedded rather than less important. A standalone chatbot or software product is visible; a model operating inside logistics, underwriting, manufacturing, fraud detection, customer service, or financial reporting may be far harder to identify.
The ONS methodology is designed to distinguish between AI production and the wider use of AI across the economy. That distinction matters because an economy can become a heavy adopter of imported or externally developed systems without necessarily building a large domestic AI supply industry.
Investment measurement presents another complication. Businesses can buy access to models as a service, build systems internally, acquire hardware, purchase software containing AI features, or pay consultants to integrate third-party tools. Each route can create economic value while appearing differently in existing datasets.
The work also provides a stronger basis for assessing labour-market claims. Estimates of jobs created, displaced, or changed by AI depend partly on how AI-intensive activities and occupations are defined. Without consistent boundaries, comparisons across studies can exaggerate differences that are partly methodological.
Government procurement is adding to the need for better measurement as public bodies test AI in healthcare, administration, defence, and other operational settings. The distinction between spending on pilots and value generated through sustained deployment becomes increasingly important as projects move beyond research funding.
The ONS has stressed that the current publication is methodological rather than a new set of economic statistics. Any headline estimate of AI’s contribution to GDP will depend on subsequent data collection, classification, and experimental work.
The value of the account will therefore emerge over time. AI is increasingly functioning as an input across industries rather than a discrete sector, and official statistics need to capture both the companies building the technology and the organisations embedding it into ordinary economic activity.
You must be logged in to post a comment.