Three Years Of AI Gave UK Firms 0.5 Per Cent, And They Expect Five Times The Rate Next

Firms surveyed by the Bank of England estimate that artificial intelligence has raised their productivity by about 0.5 per cent over the past three years, according to the central bank's November Monetary Policy Report, published on 5 November.

The Bank reads those responses, from its Decision Maker Panel, as implying that AI has added roughly 0.1 to 0.2 percentage points a year to UK productivity growth since early 2024. Over the next three years, the same firms expect AI to contribute about 0.9 percentage points a year.

The report says novel AI-related activity accounts for just under a fifth of UK GDP growth since the start of 2026, but attributes this to upstream work rather than wider adoption. The contribution to growth so far "appears to reflect activity in the upstream development and implementation of AI technologies rather than broader gains from their adoption elsewhere in the economy," it states. Productivity growth has improved in AI-exposed sectors including information and communication, but remains below pre-pandemic averages.

On employment, firms report close to zero effect so far. The Bank's regional Agents report that AI adoption is reducing the likelihood of firms hiring or replacing departing workers, but is not materially increasing the likelihood of firing them. The Bank notes that if AI raises productivity and firms reduce hiring as a result, that would represent a structural and likely persistent increase in productivity growth.

Business investment grew by nearly 5 per cent in the first half of 2026, partly on AI spending. The report accompanied a Monetary Policy Committee vote of 8 to 1 to raise Bank Rate by 0.25 percentage points to 4 per cent. The productivity and employment figures are firms' own estimates rather than measured outcomes.

Put the two survey answers next to each other and they are difficult to reconcile.

Asked what AI has done, firms said 0.5 per cent over three years, which is about 0.17 percentage points a year. Asked what it will do, the same firms said 0.9 percentage points a year for the next three. That is more than five times the rate they report achieving, forecast by people who have already spent three years trying. Expectations are not converging on experience. They are moving away from it.

The usual defence is that technologies take time, that the returns arrive after the plumbing is built, and that three years is nothing. All of that may be true. But it is worth noticing that this is the same shape of argument offered three years ago about the three years just completed, and the firms making it now are not new to the question.

What gives the 0.5 per cent unusual weight is who produced it. These are self-reported estimates from businesses that have spent money on AI and have every reason to believe it worked. Self-assessment of that kind normally biases upward, because nobody enjoys reporting that their investment did little. The number survived that bias and still came out at 0.5 per cent.

The Bank's other finding explains why the macro picture nevertheless looks strong. Just under a fifth of UK GDP growth this year is AI-related, and almost all of it is the building rather than the using. An economy can grow handsomely from constructing something whose benefits have not yet appeared, which is the position Britain is in, and the position a great many balance sheets are underwriting.

The employment finding is the one that should travel, because it describes a mechanism that statistics are built to miss.

Firms using AI are hiring fewer people and replacing fewer leavers, without firing more. That produces no redundancy announcements, no spike in claimant counts and no visible event for a labour ministry to respond to. It shows up only as jobs that were never advertised, and the people who bear it are those entering the workforce rather than those already in it.

Look at the corporate announcements of the past fortnight through that lens and they fit exactly. Barclays framed 5,000 reductions around organisational simplification and AI-driven efficiency rather than mass layoffs. Microsoft offered voluntary exit packages to about 60,000 people instead of cutting them. Both chose instruments that reduce headcount without creating a firing event. The Bank has described in survey data what those companies were doing in practice.

For India the transmission is more direct than it looks. Naukri's revenue tracks white-collar hiring volumes, and Info Edge has just put its founder back in charge of a business whose core asset is job listings. A hiring freeze is precisely the condition that a listings business cannot hedge against, because it removes the transaction rather than the customer.

The wider point is about measurement. If AI's labour effect runs through vacancies rather than dismissals, then unemployment data will look reassuring for some years while the entry-level market tightens underneath it.

For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.