TL;DR
LSE economists, two of them Treasury advisers until recently, put UK productivity growth at “an annualized productivity growth of 1.6 per cent between 2024 Q3 and 2026 Q1, compared with only 0.3 per cent over the decade preceding 2024 Q3”. The tax records underpinning that finding show employment falling by 133,000, where official statistics report a rise of 377,000. Bank of England work finds vacancies thinning fastest in the occupations most exposed to automation.
The discrepancy doing the work
Productivity is output divided by workers, so which employment series you use largely decides the answer. The researchers drew on PAYE administrative data. Had they used the official headcount instead, the same output would be spread across half a million more people and the improvement would look considerably less dramatic.
This is not a technicality. It is the difference between “Britain has started producing more” and “Britain is producing the same with fewer”, and the two imply opposite policy responses. It also lands in the same week the ONS confirmed it is deploying AI partly to repair the jobs data at the centre of this dispute — a point covered separately today.
Where the contraction concentrates
Bank of England economists find advertised roles falling away quickest in occupations where AI can substitute most readily. Customer service postings have fallen at an average annual rate of about 23% since 2023. Separate work suggests hiring is weakest in jobs that were already the lowest paid, and youth unemployment now sits at its highest in over a decade.
Recruitment itself has seized up. LinkedIn reported applications rising 45% last year, arriving at 11,000 a minute, a large share of them machine-drafted. Employers respond with longer filtering, which falls hardest on candidates with no track record to filter on.
The surrounding indicators are mixed rather than bad: unemployment reached 4.9% between April and June, vacancies fell to 707,000, and private-sector pay growth excluding bonuses ran at its slowest since the pandemic — while June output grew 0.3%, business investment rose 1.7% and the FTSE 100 extended gains to six straight quarters.
Looking forward
For UK employers the live question is whether this is a cycle or a reconfiguration. The study’s authors lean towards the latter, reading it as an early indication of what AI does to an economy rather than something confined to low-paid work.
Resultsense has tracked the same fracture from several angles: Goldman finding the pressure heaviest at entry level, Hays posting its first loss since 2003 and naming AI, Silvana Tenreyro questioning whether the technology is disinflationary at all. These figures are the first to attach a productivity number to it. The caveat is that a gain produced arithmetically by a shrinking denominator is not the same as one produced by better output — and the statistics needed to separate them are precisely the ones currently under repair.