Artificial intelligence has started appearing in Britain’s official economic statistics as more than an investment forecast or adoption survey, although the Office for National Statistics is being careful not to claim that AI itself can yet be measured as a distinct source of national growth.

The Office for National Statistics said computer programming, consultancy, and related activities grew by 3.5% in July, contributing 0.12 percentage points to real monthly GDP growth. The wider information and communication sector expanded by 2.4%.

Many of the businesses reporting the largest turnover within programming, consultancy, and information services were involved in activities related to artificial intelligence and cloud computing, according to the ONS. Its statisticians also stressed that the way the data is collected makes it difficult to quantify the exact contribution of those technologies.

That caveat prevents the July figures being described credibly as 0.12 percentage points of AI-driven GDP. The statistical category includes a much broader range of software development, consulting, systems work, and technology services, while businesses involved with AI may still earn substantial revenue from conventional products and professional services.

Even with those limits, the release provides a different form of evidence from the surveys that have dominated claims about AI’s economic impact. Instead of asking employers whether they plan to deploy AI or whether managers believe productivity has improved, the ONS is recording stronger economic activity in industries where AI and cloud work is concentrated.

The strength is not confined to one monthly comparison. Information and communication output rose 2.5% over the three months to July compared with the preceding three months, with computer programming, consultancy, and related activities among the main drivers.

That performance sits within an economy where growth remains uneven across industries. Technology-related services can expand while other sectors weaken, meaning the figures say more about the composition of growth than they do about a broad AI-driven acceleration across the economy.

Technology spending reaches measured output

New technologies normally appear in national statistics through several channels before their productivity effects become clear. Cloud companies sell more infrastructure, software developers build products, consultancies integrate systems, and businesses invest in internal development long before anyone can establish whether the resulting tools allow the customer to produce more with fewer resources.

AI appears to be following the same pattern, with the additional difficulty that it is often embedded inside products and services that already existed. A company buying an AI-enabled version of enterprise software still appears largely within conventional software expenditure, while a consultancy building an automated workflow remains classified according to the business activity rather than the specific model used.

The reverse problem also applies because companies described as AI businesses can earn money from cloud infrastructure, custom software, data engineering, or ordinary consulting alongside newer model-based services. Separating the economic effect of AI from the wider digital economy is therefore difficult even when revenue is rising quickly.

The ONS caveat reflects that measurement problem rather than an absence of activity. National accounts classify industries and outputs at a pace very different from technology marketing, while the boundaries between software, cloud, consulting, data, and AI become less distinct each time a new capability is folded into an existing service.

More persuasive evidence of a broad productivity effect would appear outside the technology suppliers themselves. Manufacturers producing more from the same equipment, logistics businesses cutting planning costs, banks reducing administrative workload, or public services processing cases faster would show the technology affecting output in industries buying AI rather than merely increasing revenue for those selling it.

That distinction is important because an investment boom can make the technology sector grow before customers demonstrate a corresponding gain. Businesses can spend heavily on cloud capacity, consultants, models, and software while implementation costs temporarily rise faster than any savings produced by the systems.

Productivity benefits can also take time because introducing software rarely redesigns an organisation by itself. Processes, responsibilities, data access, controls, and staff roles often have to change before automation reduces work rather than adding another layer of technology around the existing one.

Infrastructure spending further complicates the calculation. Data centres, computing equipment, network upgrades, construction, and energy investment can all add to economic activity while creating pressure on electricity networks, planning systems, and local infrastructure. The national contribution may therefore look positive even where regional constraints become more difficult.

July’s figures also arrive against a longer run of growth in programming activity, making subsequent releases worth watching for persistence rather than treating one month as a trend. A reversal would make the AI interpretation less persuasive, while sustained expansion alongside stronger results in customer industries would add weight to claims of a broader economic effect.

For now, the ONS has provided something narrower but more concrete than the familiar forecast about trillions of pounds of future value. AI and cloud-related businesses are visible among companies reporting strong turnover inside a part of the economy that made a measurable contribution to July output.

That still leaves a large distance between technology-sector growth and economy-wide productivity. The next stage is not proving that money is being spent around AI, but showing whether the organisations buying it ultimately produce more, operate more efficiently, or deliver better services after the implementation work is finished.