Interactive visualiser: shaune.neocities.org Github repo: https://github.com/shaunebarboutis/UK-AI-exposure-based-on-region/tree/master This paper presents a UK-specific analysis of artificial intelligence exposure across 412 occupations classified under the Standard Occupational Classification 2020 (SOC 2020) framework. Extending the methodology of Karpathy (2025), which scored 342 US occupations using a single cloud-hosted model, this study employs three architecturally distinct, locally hosted language models Nemotron Nano 4B, Qwen 3.5 9B, and Llama 3.1 8B to generate cross-validated AI exposure scores on a 0–10 scale. The analysis draws on ONS Annual Population Survey data, ASHE pay statistics, and Nomis regional employment figures to assess exposure across 31.6 million UK jobs. Key findings include: 9.9 million jobs across 94 occupations face high AI exposure, with a median pay of £40,849 compared to £28,308 for low-exposure occupations. The most exposed occupations are also among the fastest growing, suggesting AI-driven transformation rather than elimination. London’s workforce faces the highest weighted exposure (5.53) while Northern Ireland faces the lowest (4.37), with significant implications for the UK’s existing regional economic inequality. The three models achieved pairwise correlations between 0.78 and 0.85, with 75.4% of occupations scored within a two-point range, providing confidence in the cross-validated methodology. The sector rankings are broadly consistent with those in the 2023 UK government exposure analysis, and the employment patterns are consistent with the ILO’s augmentation versus substitution distinction. These results should be interpreted as measures of potential occupation reshaping and not direct evidence of current adoption or job displacement. Taken together both the inter-model agreement and external consistency checks suggest that the framework is useful for comparing relative exposure across UK occupations, while still requiring caution in causal interpretation.
Barboutis Shaune (Sun,) studied this question.