Jobs AI Is Replacing: Declining Occupations in the US and UK, With Figures to 2035

The roles shrinking fastest in the World Economic Forum (WEF), US Bureau of Labor Statistics (BLS) and UK projections share one trait: routine information processing. Data entry, typing, switchboard work, cashiering, bookkeeping and general office administration top every decline list. Generative AI sharpens that trend, though the BLS does not attribute all of these declines to AI. This guide gives the lists with figures and explains why timing a transition matters.

WEF: fastest-declining jobs worldwide, 2025 to 2030

Percentages are as reported by Euronews from the WEF Future of Jobs Report 2025.

BLS: fastest-declining US occupations, 2025 to 2035

The BLS fastest-declining table (employment in thousands, median wage May 2025) shows where percentage losses are steepest.

BLS: the largest US job losses by headcount

Large occupations lose the most jobs even at modest percentages. The BLS table of largest job declines lists:

By occupational group, office and administrative support is projected to fall 4.0%, or 752,100 jobs, the largest loss of any group in the 27 August 2026 release. Sales and related occupations fall 1.4% and production 0.4%. Total US employment still grows 3.5%.

UK evidence

The UK sources differ in age. The ONS analysis uses 2017 data, the DfE report appeared in 2023 and the NFER release in November 2025. None separates AI from other automation, so use them to judge direction and compare the newest figures first.

Exposure and decline are different measures

The lists disagree in useful ways. The DfE ranks accountants among the most AI-exposed UK occupations. The BLS projects US accountants and auditors to grow 5.0% (79,400 jobs), while bookkeeping clerks shrink 5.6% (85,600 jobs). Exposure shows how many tasks AI can assist with. Decline shows what employers plan to do about it. Where AI augments a skilled worker, headcount can hold. Where it replaces a routine task bundle, headcount falls.

The BLS release shows the same split inside one economy. Office and administrative support falls 4.0% while computer and mathematical occupations grow 7.3% and data scientists grow 34.6%. The BLS adds that its exposure tiers do not forecast job loss and do not separate automation from augmentation.

Research estimates vary widely with method. Frey and Osborne (Oxford Martin School, 2013) put 47% of US employment at risk of computerisation. A study by Eloundou, Manning, Mishkin and Rock (arXiv, 2023) found about 80% of US workers could have at least 10% of their tasks affected by language models and about 19% at least half. The ONS found 7.4% of English jobs at high risk. Treat each as a different question.

What the declining roles have in common

The OECD AI exposure measure (May 2026) finds the occupations closest to current AI capability involve routine information processing, administrative work and codifiable tasks. The furthest are those requiring contextual judgement, interpersonal understanding, complex decision-making and responsibility. The decline tables above match that pattern: typists, keyers, clerks and operators handle structured inputs with rules.

Three practical readings follow. First, a civilian title in records, personnel administration or customer processing will appear on these lists, so check the title before you apply. Second, security and public-service roles carry their own pressures: the BLS projects correctional officers down 7.4%. Third, the pay in declining roles often sits at or below the all-occupations US median of $50,980, which makes them tempting fallback jobs. Treat them as a bridge with an end date.

Why timing matters

The WEF estimates that 59 in every 100 workers need training by 2030, and that 11 of them are unlikely to receive it (skills outlook). Employers split on the response (workforce strategies):

• 85% plan to upskill existing staff and 77% plan reskilling to work alongside AI
• 47% plan to move staff out of AI-disrupted roles
• 41% expect to reduce headcount as AI replicates tasks

You cannot tell in advance which response your employer will choose. A short audit before you sign a lease or enrol on a course costs little. The same audit two years into a shrinking role costs far more.

A five-minute audit of your target role

1. Search the role in the BLS tables above and in the WEF lists. Note the percentage and the headcount change.
2. Open the BLS AI exposure file for your occupation and note its tier. Treat the tier as a prompt for questions.
3. Ask the employer which tasks in the job AI tools already handle, and how the team will change.
4. Look for the adjacent role that supervises, checks or sells the work. Customer service teams, for example, still need quality, escalation and team leader roles.
5. Pair your plan with a growth list. Our guide to the WEF 2030 jobs forecast lists the roles expanding fastest.