TL;DR

Bank of England staff have found early evidence that AI is feeding into UK productivity growth, but almost all of it sits in the sectors that build AI rather than the ones using it. Computer programming and consultancy multiplied its share of yearly productivity growth by ten across the two periods compared, moving from 0.01 percentage points to 0.10. Information services flipped from dragging on growth to adding 0.06 points.

The analysis, by Sandra Batten, who works in the Bank’s structural economics team, uses ONS output-per-hour data and was carried out in the ONS Secure Research Service. It treats generative AI as a general purpose technology — pervasive, continuously improving, and capable of spawning further innovation — and tests whether it is following the path of earlier ones.

That path has two phases. During the computing revolution, gains showed up first among the firms making semiconductors, hardware and software, and only later among the industries buying them. Britain appears to be in phase one.

Where the using sectors stand

Among AI adopters rather than producers, administrative and business support services made the strongest showing, having previously dragged on growth. Manufacturing also contributed. Financial and insurance activities did not: despite being a large sector and a heavy adopter, its contribution was negative, which Batten attributes to possible output mismeasurement or other pressures cancelling out AI’s effect.

Plotting industries by adoption rate against their change in productivity contribution produces an upward-sloping line, though a weak one — the fit explains around a tenth of the variation, and the paper is explicit that this is correlation, not causation.

Looking forward

For UK businesses the finding is a useful corrective in both directions. It undercuts the claim that AI has delivered nothing measurable, and it undercuts vendor promises of immediate returns for buyers. If the computing analogy holds, adopter-side gains arrive only after adoption becomes widespread and firms reorganise around the technology. The finance sector’s negative contribution is the number worth watching: the UK’s most AI-invested industry currently has the least to show for it.