How AI Accounted for Nearly a Fifth of UK Growth This Year,

5 Nov

AI Is Lifting UK Growth. The Productivity Question Is Harder to Answer — When Will Its Customers See a Return?

Britain Can Earn Money From AI Before A Single Customer Saves An Hour.

A British software firm can spend six months building an AI system for a client. During those six months, its developers have produced something of value in the UK. The client may still be waiting to find out whether the system saves enough time to justify the bill. Both things can be true, and the gap between them explains one of the most striking figures in the Bank of England’s November report.

On 5 November, the Bank published an estimate that new AI-related activity accounted for just under a fifth of UK GDP growth since the start of 2026. Its economists found the clearest evidence among businesses developing software, advising customers and helping them put AI systems to work. The evidence for a productivity boost across the businesses buying those systems is thinner.

What Does Nearly A Fifth Of Growth Mean?

The figure describes a share of the increase in British economic output. Imagine an economy producing £100 of goods and services in one period and £101 in the next, measured at constant prices. If AI-related activity accounted for just under a fifth of that growth, its contribution would be a little under 20p of the extra £1. The £100 of existing output sits outside that calculation.

That arithmetic helps explain how activity concentrated in a few industries can produce such a large headline. If output elsewhere changes little, an unusually busy year for software developers can account for a sizeable share of the national increase. The measure also works in reverse. If the rest of the economy accelerates, AI’s share of growth could fall even as British AI businesses continue to expand.

The period matters too. The Bank was looking at growth since the beginning of 2026 using the evidence available in November. A final figure for the full calendar year will have to wait for the remaining months of economic data.

How Did The Bank Find AI In The GDP Figures?

A company that writes software for hospitals might spend Monday improving a conventional booking system and Tuesday developing an AI tool that reads referral letters. Both jobs can appear under computer programming. The official industry label tells economists what sort of company did the work. It rarely tells them how much of the week went on AI.

Bank staff tackled that problem by combining output from three groups where they would expect AI work to show up. Computer programming and consultancy covers work such as advising a business on which model to use. Information services includes data processing and data-centre services. The less obvious group is publishing. In the economic statistics, a publisher can be a business releasing application software, including an AI agent.

There was a reason to look closely at those industries. In the three months to August, UK output from computer programming, consultancy and related activities rose 5% compared with the preceding three months. Across the broader information and communication sector, the increase was 2.8%. The Office for National Statistics also found that AI and cloud computing featured among businesses reporting particularly strong turnover growth.

The Bank then made the adjustment on which the headline depends. British technology businesses were expanding before the latest wave of AI spending. Counting all their growth would credit AI with work that might have happened anyway. Staff allowed for that longer-running trend and used the additional activity to estimate AI’s contribution.

There is room for error in both directions. A surge in demand for conventional software could find its way into the estimate. An AI tool developed entirely inside a retailer could escape it. The underlying output is observable. Deciding how much of the acceleration belongs to AI requires judgement.

Where Does The Money Enter The British Economy?

Consider a hypothetical UK insurer commissioning a system to sort incoming claims. It buys access to an overseas AI model and hires a British consultancy to connect that model to its claims database. The consultants have to decide which documents the system can read, test its handling of ambiguous cases and give staff a way to correct errors.

The distinction that matters for UK GDP is where the value is created. The British consultancy produces a service here. The fee for a model supplied from overseas buys an import. Both appear in the insurer’s project budget, but adding up the invoices and calling the total a contribution to British GDP would overstate what Britain produced.

The same principle applies to a data centre. A server built abroad can be expensive, while the work of installing, configuring and operating it in Britain generates domestic activity. A headline investment announcement gives the size of a proposed budget. To understand its contribution to GDP, an economist needs to know what was delivered, when and where.

The Bank found supporting evidence in investment figures. Total UK business investment grew by nearly 5% in the first half of 2026, with strong growth in information and communication. Its business contacts also identified AI spending as a factor. Utilities investment contributed to the wider rise as well. The investment figures give a second view of the spending boom, and some of the same software work can appear in both views. Adding the two figures together would count part of the activity twice.

When Does An AI Purchase Become A Productivity Gain?

Return to the insurer. Before the new system arrived, a claims handler read each submission and sent it to the right team. Afterwards, the software makes the first decision. A stopwatch might show that the initial sorting takes seconds. The insurer still needs to know what happens to the claim.

If the software sends difficult cases to the wrong team, the time saved at the start can disappear into corrections, repeat calls and complaints. If it routes cases accurately and staff can settle more claims to the same standard in the same number of hours, the insurer has evidence of a productivity gain. The people best placed to establish that are the ones who can follow a claim all the way through.

That distinction also matters to the Bank of England. A rush of orders for developers increases demand for their services now. A lasting improvement in what insurers, manufacturers or other customers can produce in an hour would expand the economy’s capacity over time. Those developments can arrive years apart, and they have different implications for an institution trying to judge how fast Britain can grow without adding to inflation.

The most revealing evidence may eventually come from a company that has stopped talking about its AI project altogether. Once the installation bill has been paid and the consultants have left, can the same team settle more claims accurately, month after month? Britain already has evidence that it can earn money building these systems. The claims ledger is where the customer’s return will show up.