What the tool is, and is not
It is a map of how much of each UK occupation's work a language model can do today. The unit is the SOC 2020 four-digit occupation group. Employment comes from ONS ASHE, and the exposure scores come from Eloundou et al's 2023 paper.
It is not a prediction that any role disappears. It is a snapshot of where AI has already arrived. The more useful question is which tasks get automated, which get augmented and which stay human. The treemap starts that conversation rather than ending it.
Three readings of the same data
The task scores come with three ways of adding up. Alpha counts tasks the model can do on its own. Beta gives partial credit when the model is paired with tools. Gamma counts any exposure, direct or with tools.
- Alpha: £107.2bn of UK wages. The strict reading: can a model do this task unaided?
- Beta: £332.75bn. The middle case, with partial credit where model plus tools is the real story.
- Gamma: £558.32bn. The most generous reading, and the closest to how AI is used at work in 2026.
The tool's front page leads with gamma, because that matches the shape of the paper's own US figure of $711bn, and because tools are normal now. The tiles and each occupation's score use alpha, which keeps the individual readings conservative. Both are in the data, and each surface picks one.
How occupations meet task scores
UK SOC 2020 unit groups are bridged to O*NET tasks through a chain of public crosswalks.
- The ONS SOC 2020 coding index (about 32,000 job titles) gives the most common ISCO-08 code for each UK unit group.
- The US BLS crosswalk maps each ISCO-08 code to one or more US SOC 2010 codes.
- O*NET-SOC codes cut to six digits match US SOC 2010.
- Task scores are averaged up that chain: weighted by importance within an O*NET-SOC code, then plain averages back up through SOC 2010 and ISCO-08 to UK SOC 2020.
The chain loses coverage at every step. In the end 379 of 412 UK unit groups have a score (91.99%). Of those, 326 also have ASHE employment data, so 299 can be scored for wage exposure.
The limits
- The task labels date from 2023 and were made against early GPT-4. Current models can do considerably more, so the figures are likely to understate today's exposure.
- ASHE counts employees only. Self-employed workers are missing, so roles with a large freelance share (creative work, consulting, trades) are under-weighted in the wage bill.
- The ISCO-08 bridge is lossy. The coding index gives several ISCO codes per UK group, and the tool takes the most common one. Other choices give slightly different numbers.
- Exposure is not the same as a job going away. It is a ceiling on how much of the task list a model can touch. How much of that is automated, augmented or kept human is a design choice, not a technical one.
Sources
- ONS SOC 2020 Volume 1, structure and descriptions of unit groups, version 2025-12-03, OGL v3.0. Office for National Statistics (2025). Standard Occupational Classification 2020 (SOC 2020), Volume 1: Structure and descriptions of unit groups, revised 3 December 2025. Licensed under the Open Government Licence v3.0.
- ONS SOC 2020 Volume 2, coding index (contains SOC 2020 ↔ ISCO-08), version 2025-12-03, OGL v3.0. Office for National Statistics (2025). Standard Occupational Classification 2020 (SOC 2020), Volume 2: The coding index, revised 3 December 2025. Licensed under the Open Government Licence v3.0.
- O*NET 30.2 database (tab-delimited, ZIP), version 30.2, CC BY 4.0. National Center for O*NET Development. O*NET 30.2 Database. Employment and Training Administration, U.S. Department of Labor. Used under CC BY 4.0.
- Eloundou et al, per-task AI exposure scores (full_labelset.tsv, alpha = direct exposure), version 0471612f (pinned), MIT (repo). Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2024). GPTs are GPTs: Labor market impact potential of LLMs. Data: github.com/openai/GPTs-are-GPTs (commit 0471612f).
- US BLS, ISCO-08 ↔ US SOC 2010 crosswalk, version 2012 (crosswalk current), Public Domain (US Government work). U.S. Bureau of Labor Statistics (2012). ISCO-08 to 2010 SOC Crosswalk. Public domain.
- ONS ASHE Table 14, occupation (4-digit SOC 2020), 2025 provisional (includes employment count and gross annual pay), version 2025 provisional (released 2025-10-23), OGL v3.0. Office for National Statistics (2025). Annual Survey of Hours and Earnings (ASHE), Table 14: Occupation (4-digit SOC 2020), 2025 provisional. Licensed under the Open Government Licence v3.0.
The full write-up
Source-by-source notes, the formulas, the coverage lost at each join and the code fragments are all in the technical PDF, generated from the same data as this page.
Spotted an occupation that is badly mapped, a task score that does not match your day, or a choice you would make differently? Email hello@alexlockey.com and the better calls go into the next build.
