UK AI JobsExpected · in use · happened
What AI does · Anthropic Economic Index

A writing machine. Not a coding one.

Forget the forecasts. Anthropic publishes what millions of real Claude conversations actually do, job by job. Weighted across the whole UK workforce, the three things it most often produces are documents, explanations and advice: 43% of all measured output. Everything a developer would call building software, the apps, scripts, fixes and queries, adds up to about 6%.

What AI makes at work · job-weighted share of measured output
Documents & reports
16.7%
Explanations & answers
15.0%
Advice & recommendations
11.7%
Analysis & summaries
6.2%
Plans & strategies
5.8%
Spreadsheets & data
5.5%
Emails & messages
4.1%
Teaching material
2.9%
Apps & websites
2.2%
Recipes & meal plans
2.1%
Charts & visuals
1.8%
Maths & calculations
1.7%
knowledge outputbuilding software23.4M workers · shares ≥2% sum to ~84%
Assist or replace · the split

Assist the person, or do the task? Almost dead even.

Cross the measured usage with ONS employment and the workforce splits down the middle. Of 23.4M mapped workers, 12.3M are in occupations where usage leans toward doing the task, and 11.1M where it leans toward helping the person. “Leans” is a majority-share cut of how AI is used in a job, not a count of jobs lost.

12.3M · automation-leaning
augmentation-leaning · 11.1M
7.1×
It is genuinely capable: a task that takes a person about 4.3h alone drops to roughly 36 min with AI. Model-estimated times.
12.2 yr
And it works at graduate level, at or above what the role itself needs in 98% of UK jobs. Capable, yet the headcount cuts have not shown. Model-estimated.
How you work with it · the five measured modes

How that near-even split actually happens.

Every Claude conversation gets tagged with how the person and the model worked together. The single most common mode is directive: hand it a task and take the output, which counts as automation. But the three assisting modes, iterating, learning and checking, together nearly match it. That texture is why the split lands near even: 45.7% of collaboration assists the person, 50.3% does the task.

Directive
hand it a task, take the output
35%
Task iteration
refine it back and forth
28%
Feedback loop
fix what the AI got wrong
15%
Learning
ask it to teach the domain
15%
Validation
check work that is already done
3%
assists the person (augmentation)does the task (automation)a further 4% is unclassified
Who gets which half · by pay

The lower the pay, the more the AI just does the task.

Split the occupations into pay fifths and the split stops being even. Automation-style usage runs 57.9% in the lowest-paid fifth and falls, step by step, to 49.3% at the top, while the assisting share climbs from 42.1% to 50.7%. Automate the poor, augment the rich. The tilt is real, and it is gentle.

Automation falls as pay rises, augmentation climbs · they cross just below the top fifth
Automation-style vs augmentation-style share of usage by pay fifth40%45%50%55%60%£2027kwork 33%£2732kwork 47%£3236kwork 49%£3643kwork 50%£43154kwork 51%57.9%42.1%49.3% automation50.7% augmentationthe crossing
Axis runs 40–60%; the 50% line is marked. Carry the work row with you: in the lowest-paid fifth only ~33% of measured usage is work at all (the rest is personal), against ~51% at the top, so the left of the chart leans on thinner work evidence. The full wallet cut →
The age myth · measured

The bottom rung isn't where the fear says it is.

The fear says AI eats the young's tasks first. Measured, automation-style usage is flat across every age band: between 51.7% and 53.8% from 16 to 65+. What is not flat is where the autonomy sits. Only 14.1% of 16–24 employment is in occupations in the top quarter for AI autonomy, against 25% for every other band: 44% under-represented.

Two measures per band, one true 0–60 scale · the tasks are flat, the autonomy is gated
0%20%40%60%
auto · autonomy
16-24
53.8% · 14.1%
25-34
51.7% · 23.8%
35-44
51.7% · 25.3%
45-54
52.2% · 24.8%
55-64
52.9% · 23.9%
65+
53.1% · 27.1%
automation-style share of the band's usageshare of the band's jobs in top-quartile-autonomy occupations

AI isn't doing young people's jobs. It may be doing the jobs young people would have been hired to learn. The squeeze shows up at the hiring gate, in entry-level postings and first jobs, and the realised side of that story is on the gap tracker.

AEI usage per occupation crossed with age-band employment (ONS APS ad-hoc 3410, 2025) on the dual-coded SOC crosswalk; coverage 92–96% of each band. How this is built →

Britain vs the world · 121 countries

One of the heaviest AI adopters on earth, and one of the most collaborative in how it uses it.

17 of 121
Adoption rank
Claude usage per working-age person
5th
Percentile · automation-style share
nearly every other country automates more
19th
Percentile · AI autonomy
UK usage keeps the human in the loop

Two caveats, carried openly. Rich, high-adoption countries skew toward augmentation-style usage, so part of this is who Britain's users are rather than a national virtue. And this strip is the one genuinely UK usage cut on the site: every per-occupation number elsewhere is a global usage fingerprint applied to UK employment weights. Said plainly in the method →

Which jobs · the 12 biggest occupations

The same split, across the jobs most people actually do.

Kitchen, warehouse, care and admin work lean toward automation-style use. Teaching, finance and sales lean toward augmentation. Software developers sit close to the middle. The wave, when it comes, will not land evenly.

Kitchen and catering assistants
29%
Warehouse operatives
31%
Other administrative occupations n.e.c.
34%
Care workers and home carers
36%
Programmers and software development professionals
43%
Book-keepers, payroll managers and wages clerks
49%
Nursing auxiliaries and assistants
50%
Sales accounts and business development managers
56%
Financial managers and directors
58%
Sales and retail assistants
59%
Secondary education teaching professionals
63%
Primary education teaching professionals
67%
augmentation (AI assists)automation (AI does the task)See all 738 on the map →

Measured: Anthropic Economic Index, real Claude.ai usage, Apr+May 2026, global, CC-BY. Confounds: Claude users only, not the whole labour market; the US O*NET-SOC to UK SOC-2020 crosswalk drops ~9% of codes; only ~45% of mapped usage is work, not personal; two-month cross-section. Employment: ONS. How this is built →

Next

Now find where it would land first.