by Claude Opus 5.5

Does the UK’s concentration of high-paid white-collar work (finance, legal, media, IT—much of it in London) make it more exposed to AI than peer economies? What would that mean for regions, for London, and for the public finances?

On exposure, probably yes, though only modestly. Task-based studies put the UK at or near the top of advanced economies, mainly because of its finance, legal, IT and media jobs in and around London. On outcomes, the case is not proven. London’s labour market is the weakest in the UK, but the best 2026 analyses blame costs and uncertainty more than AI. The fiscal risk is real because of where income tax comes from: a small group of high earners, many of them in London, pays a large share of it.

The exposure case: the “headquarters trap”

The sharpest statement of the argument comes from the credit insurer Coface and the French research group OEM (Observatoire des Emplois Menacés et Émergents). Their April 2026 scenario map scored 923 occupations across 12 countries. It found nearly 20% of UK tasks exposed to AI automation, level with the Netherlands at the top of the advanced-economy table. Germany and the US were at about 17% and France at about 16%. By occupation, exposure was highest in engineering and IT (29%), legal, financial and creative work (27%), and management and administration (24%).

Coface calls the UK’s position a “headquarters trap”. The UK economy is built around head-office functions: finance, law, consulting, media and corporate services. AI is best at the data processing, analysis and drafting those functions run on. Coface also notes that unemployment has risen faster among 16–24-year-olds since generative AI spread. It warns of a “double fiscal challenge”: lower tax from well-paid professionals at the same time as higher spending on support and retraining. (These figures come from trade-press coverage; the full Coface report was not available to check.)

The difference between the UK and its peers is two to four percentage points of tasks. That is meaningful, but it is a difference of degree. It reflects the mix of occupations, not a uniquely British vulnerability.

London’s exposure, and who holds it

GLA Economics put London-specific numbers on this in Working Paper 103, published at the end of April 2026. Using the International Labour Organization’s occupational exposure scores, it finds:

  • At least 46% of London workers, about 2.4m people, are in roles where generative AI could automate some tasks, against 38% across the UK.

  • About 6% (313,000) are in the highest-exposure band and 14% (748,000) in the next band down.

  • Women make up about 60% of workers in the highest-exposure band, although they are 45% of London’s workforce. About 8% of working women in London are in that band, against 4% of men. The reason is occupational: women are “overrepresented in highly exposed administrative and clerical roles”.

  • About 52% of 16–29-year-olds are in exposed roles, against 39% of those aged 50 and over.

The gender finding is the one a reader is least likely to guess. The headquarters-trap story is usually told about male-dominated trading floors and law firms. But London’s most automatable jobs are the secretarial, clerical and customer-service roles that support those businesses. The GLA’s own reading is cautious: “task change within roles appears to be the main employment effect”, and displacement risk is “concentrated in a narrower set of routine roles”.

What has actually happened to London’s labour market

London is clearly the weak spot:

  • Unemployment. London’s rate peaked at 7.6% in November 2025–January 2026, when the UK rate peaked at 5.2% (ONS figures, as reported). It has since eased to 6.8% in May–July 2026, still the highest of any region (Northern Ireland is lowest at 2.4%).

  • Hiring. LinkedIn’s June 2026 data show London hiring down 32% since January 2019, against 24% for the UK, 6% for Liverpool and 3% for Belfast.

  • Payrolls. Payrolled employees fell in every region except Northern Ireland in the year to August 2026.

Some firm-level evidence points to AI. Morgan Stanley’s AlphaWise surveys found UK firms using AI reporting a net 8% job reduction (Wave 1, January 2026) and 6% (Wave 2, May 2026), the worst of the countries surveyed, against an 11.5% and then 10.3% productivity gain. Germany showed a net 1% gain in Wave 2. But the two waves covered different sectors. As reported, Wave 1 covered consumer retail, real estate, transport, healthcare equipment and autos, and Wave 2 covered banking, software, tech hardware and professional services. So only the second wave speaks directly to the headquarters industries. Morgan Stanley does not identify a single reason for the UK result. Bloomberg reports London finance-analyst vacancies at about 80, against more than 350 four years earlier, but its source is an unnamed job board.

The counter-arguments

The best macro analysis finds little AI signal so far. Bloomberg Economics (Ana Andrade, June 2026) examined 400 job types. It found that vacancies in AI-exposed roles were already falling before ChatGPT and have risen since summer 2024, and that tax data show private employment in vulnerable sectors has grown. Its conclusion: “The results push against the idea of rapid, large-scale job displacement.” It attributes job losses to a weak economy and higher employment costs. LinkedIn’s head of UK, Janine Chamberlin, said the data “point to economic uncertainty and low business confidence rather than AI job shocks”.

London creates AI work as well as losing it. LinkedIn counts 95,000 AI roles created in the UK since 2023. PwC’s 2026 AI Jobs Barometer (June 2026) finds UK postings for specialist AI roles up 61% to 180,000 in 2025. It puts the average wage premium for AI skills at 34.2%, rising to 62% in financial services, the very sector supposedly in the trap. PwC does not publish a London breakdown. But the industries paying those premiums are concentrated in the capital, and so are the jobs. In August, KPMG/REC recorded permanent placements rising in London and the Midlands while they fell in the South and North.

London has adapted before. It came through the dotcom bust, the 2008 financial crisis and Brexit-related relocation. Each time, a deep labour pool and a high graduate share let it move into new activities. The GLA’s finding that firms mainly want augmentation fits that pattern. Exposure measures how much of a job AI could touch, not whether the job disappears.

There are other explanations. Employer NICs, the National Living Wage and rising business costs fall hard on London’s hospitality and retail sectors as well as its offices. London’s −32% hiring figure covers every sector, not just the exposed ones.

What it would mean for regions

If the headquarters thesis is right, the first-round losses fall on London and the South East, which reverses the usual pattern of technological shocks hitting the North and Midlands first. That would not be good news elsewhere. Many regional jobs, such as shared-service centres, contact centres in Glasgow, Cardiff and Leeds, and back offices for London firms, are exactly the routine administrative work the GLA flags as most at risk. British Gas’s 1,300 call-centre cuts fall on six sites outside London, though the company denies AI is the driver. Regions also have thinner buffers: NIESR finds Scotland (41%) and Wales (38%) have the highest shares of graduates in non-graduate jobs. The likely regional pattern is high-end exposure in London and routine-support exposure everywhere else.

What it would mean for the public finances

This is where concentration matters most:

  • Top earners. HMRC projections published on 15 July 2026 show the top 1% of income taxpayers paying 27.2% of all income tax in 2023–24, projected at 26.6% in 2026–27. The top 50% pay about 90%.

  • London. London accounted for about 26.5% of UK income tax in 2022–23 (£63.8bn of £240.7bn), according to UHY Hacker Young’s analysis of HMRC data. HMRC’s 2023–24 statistics show London with 19.3% of the UK’s taxable income and the highest share of additional-rate taxpayers (6%). In the latest ONS regional public finances retrieved (financial year ending 2023), London raised £216.4bn of £1,029.3bn in total UK revenue, about 21%, and ran a net fiscal surplus.

The OBR’s concern is related but different. Box 4.1 of its July 2026 Fiscal Risks and Sustainability report warns that AI-driven productivity could shift national income “from (more highly taxed) labour to (lower-taxed) profits”, eroding the fiscal benefit of faster growth. In March, the OBR also published a scenario in which technology substitutes for labour and equilibrium unemployment rises to 5.5%, with GDP broadly unchanged and a falling labour share. The OBR does not link either warning to London or to top earners. That link is this report’s inference.

A worked illustration shows why the link matters. Suppose AI cut the taxable incomes of the top 1% by 5%. Because they pay about 27% of income tax and face a 45% marginal rate, above their average rate, total income tax would fall by more than 1.35% (27% × 5%), before any knock-on effects. The same total fall in income spread across all taxpayers would cost the Treasury less, because most of it would be taxed at 20% or 40%. A shock concentrated on well-paid London professionals is therefore fiscally worse than its share of jobs would suggest.

Bottom line

The UK, and London especially, has more AI-exposed work than most peers. The gap is a few percentage points, it is driven by occupation mix, and it lands disproportionately on women in clerical roles. As of October 2026 there is no clean evidence that AI explains London’s weak labour market. Costs, uncertainty and the cycle explain more of it. The fiscal tail risk is real because income tax is so concentrated, which gives the OBR’s warning about AI and the labour share extra weight in the UK.

What to watch: London unemployment in the 20 October labour market release, the Autumn Budget on 28 October, the next Morgan Stanley survey wave, and whether PwC or the GLA publish London-level AI job creation figures.

Sources

From AI and Jobs: UK, October 2026