The UK government spent almost two years citing a £400 billion potential economic boost from AI. Ministers have now set that figure beside a much lower estimate: a central projection of £120 billion by 2030, which the AI minister described as highly uncertain. The difference raises questions about how estimates are sourced, updated and used in public claims about technology.
Two estimates, different assumptions
AI minister Kanishka Narayan told Parliament that AI adoption and data investment could add between £55 billion and £140 billion to the UK economy by 2030. The central estimate is £120 billion. Narayan called this more recent assessment a “hugely uncertain” estimate.
The £400 billion figure had a different origin. Narayan said it came from a publicly available 2023 report commissioned by Google and produced by Public First, a British consultancy now referred to as PF Global. The government’s newer estimate is less than a third of that headline figure.
PF Global’s head of economics, Jonathan Dupont, said the gap partly reflects assumptions about how quickly AI would be adopted. In comments to WIRED, he said the government’s numbers assume slower adoption than seemed possible when the consultancy built its 2023 model. As more is learned about a fast-moving technology, the assumptions behind projections may need to change.
The older number remained in circulation
Government departments and agencies used the £400 billion figure for almost two years, across two governments. Ministers had the lower estimate since at least November 2025, according to the article, but Skills England was still publishing the older figure four months later.
Narayan’s parliamentary answer was the first recorded evidence WIRED could find of the government placing the two estimates side by side and identifying the report’s Google commission. The answer did not withdraw the £400 billion figure or state that it had been wrong. That leaves an important distinction: a revised projection can change the picture without the original claim being formally retracted.
The timing has drawn criticism from lawmakers. Lord Holmes of Richmond said the government should have clarified the change sooner, after the figure had been repeated by ministers and other organizations. He also pointed to the importance of robust data when public spending decisions depend on it.
Forecasts depend on uptake
PF Global said it still sees significant potential in AI, but its central estimate of the gains likely to be achieved by 2030 is now closer to the government’s figure, given the slower adoption seen in practice. The consultancy said growth could accelerate in the years after 2030.
It also pointed to ways adoption might increase, including stronger direction from the top and deeper integration of AI into civil service workflows. These are possible drivers of future gains, not evidence that the higher forecast will be realized. The estimates depend in part on what organizations do with the technology and how quickly they put it into use.
That makes a single large headline number difficult to interpret on its own. It can describe potential under certain assumptions, while leaving unclear how much of that potential is likely to materialize by a particular date. The lower government estimate and PF Global’s updated view both underline the role of adoption assumptions in estimating AI’s economic impact.
Transparency matters as much as potential
Dame Chi Onwurah, who chairs the Commons Science, Innovation and Technology Committee, said the government is right to promote technology’s potential to improve public services. She also warned that overstating benefits can weaken support for change and called for realistic claims and transparency about possible savings.
Sumedha Deshmukh of the Ada Lovelace Institute raised a related concern: precise-looking AI figures can contain assumptions and incentives, and industry may fund or shape the research behind them. She argued that governments should test figures before using them to shape policy, support independent research and track whether promised gains appear.
PF Global said its work for Google continued, while maintaining that its research was independent and its judgments and assumptions were its own. The article says the Cabinet Office and Google were approached for comment.
The dispute is ultimately about more than which estimate is larger. When governments use forecasts to explain AI’s economic promise, the source, assumptions, time frame and expected beneficiaries all help the public assess what the number means. Clearer disclosure can help distinguish a projection of possible gains from a reliable account of what has been achieved.