UK AI Job Vulnerability Dashboard

Analysis of UK occupations exposed to AI automation

Key Statistics

Overview of UK workforce exposure to AI-driven job vulnerability, based on 2024 ONS employment data and GPT task exposure analysis from Eloundou et al. (2024). High vulnerability is defined as a score above 50%. Note: The "at risk" figure should be interpreted with caution due to occupation mapping uncertainty and data imputation. *Self-employed workers (~4M) are excluded from the analysis as ASHE only covers employees.

--
Total UK Workers Analysed*
--
Workers in High-Vulnerability Jobs
--
% of Workforce at High Risk
--
Occupations Analysed

AI Exposure vs Adaptive Capacity

Each point represents a UK occupation. Size indicates employment count. Colour indicates vulnerability (red = high exposure + low adaptability). Occupations in the bottom-right quadrant (high exposure, low adaptability) face the greatest displacement risk. Hover over points for detailed occupation information.

Most Vulnerable Occupations

The six occupations with the highest vulnerability scores. These roles combine high AI task exposure (many job tasks can be performed or augmented by LLMs) with low adaptive capacity (older workforce, geographically dispersed, limited retraining options). Vulnerability = sqrt((1 - adaptability) × exposure).

All Occupations

Complete list of UK occupations (SOC 2020 4-digit codes) with AI exposure scores, adaptive capacity index, and calculated vulnerability. 368 of 412 occupations are included; 44 are excluded due to ONS data suppression. Use the search and sort controls to explore specific occupations or sectors. * = low confidence match (exposure score may be inaccurate)

Occupation Employment AI Exposure Adaptability Vulnerability

Data Sources

GPT Exposure Scores
Eloundou et al. (2024) "GPTs are GPTs: Labor Market Impact Potential of LLMs"
Science, Vol. 384, Issue 6702, pp. 1306-1308
DOI: 10.1126/science.adj0998
Data: github.com/openai/gpts-are-gpts
UK Employment Data
ONS Annual Survey of Hours and Earnings (ASHE) Table 14
2024 Provisional Release (October 2024)
ons.gov.uk/...datasets/occupation4digitsoc2010ashetable14
Coverage: United Kingdom, 4-digit SOC 2020 occupations
Occupation Classification
ONS Standard Occupational Classification 2020 (SOC 2020)
412 unit groups at 4-digit level
ons.gov.uk/.../soc2020
Regional Employment Data
ONS Annual Survey of Hours and Earnings (ASHE) Table 15
2024 Revised Release (October 2024)
ons.gov.uk/...datasets/regionbyoccupation4digitsoc2010ashetable15
Coverage: Employment by UK NUTS1 region × 4-digit SOC 2020
Used for: Geographic density component of Adaptive Capacity Index
Age Distribution Data
ONS Annual Survey of Hours and Earnings (ASHE) Table 20
2024 Revised Release (October 2024)
ons.gov.uk/...datasets/agegroupbyoccupation2digitsocashetable20
Coverage: Employment by age group × 2-digit SOC 2020
Used for: Age (55+) component of Adaptive Capacity Index
Note: 55+ approximated as 50% of 50-59 bin + 100% of 60+ bin

Methodology

US to UK Occupation Mapping
US O*NET-SOC codes mapped to UK SOC 2020 using a three-tier approach:
  1. Manual Overrides: For occupations where terminology differs significantly (e.g., UK "Physiotherapists" → US "Physical Therapists"), manual mappings are used. See full mapping table below.
  2. Synonym Substitution: UK terms are translated to US equivalents (e.g., "centre" → "center", "colour" → "color") before matching.
  3. Fuzzy Title Matching: Jaccard word overlap similarity for remaining occupations.
This approach achieves 100% coverage of UK occupations.
AI Exposure Measure
Uses the "beta" measure from Eloundou et al.: weighted sum of tasks exposed to GPT-4 (E1 + 0.5*E2), where E1 = directly exposed tasks and E2 = tasks exposed with tools.
Adaptability Index
Calculated using UK-adapted version of Manning & Aguirre (2026) methodology with three components:
  • Geographic Density: Employment-weighted log(workers/km²) across UK regions. Higher density indicates more urban concentration, providing more job options for displaced workers. Source: ASHE Table 15.
  • Age (55+): Fraction of workers aged 55+ in occupation (reversed). Higher fraction of older workers indicates lower adaptability due to shorter remaining careers and potential retraining barriers. Approximated from ASHE Table 20 age bins. Source: ASHE Table 20.
  • Skill Level: ONS SOC 2020 skill classification (1-4 scale). Higher skill level indicates more transferable skills and easier retraining. Level 4 = degree-level (SOC 1-2), Level 1 = no formal qualifications (SOC 8-9). Source: SOC 2020 major group.
Process: Winsorize at 5th/95th percentile → Z-score normalize → Average (1/3 each) → Percentile rank (0-1).
Note: 3 of 4 original components available; Net Liquid Wealth would require UK Wealth and Assets Survey data.
Vulnerability Score
Calculated as: sqrt((1 - adaptability) * exposure)
Following the methodology in Manning & Aguirre (2026) NBER Working Paper w34705.

Limitations

  • 368 of 412 UK SOC 2020 occupations included (44 excluded due to ONS data suppression where sample sizes are too small for reliable estimates, or zero recorded employment)
  • 3 of 4 Manning & Aguirre adaptability components available for UK (missing: Net Liquid Wealth)
  • Age data available only at 2-digit SOC level, applied uniformly to 4-digit occupations within each major group
  • Fraction 55+ approximated from age bins (50% of 50-59 + 100% of 60+)
  • Occupation mapping between US O*NET-SOC and UK SOC 2020 introduces some uncertainty
  • Employment figures are point-in-time (2024) estimates and may not reflect recent changes
  • Analysis focuses on task exposure potential, not actual job vulnerability rates or timing

UK → US Occupation Manual Overrides

For occupations where UK and US terminology differs significantly, manual mappings are used to transfer AI exposure scores. These 49 mappings cover occupations representing over 3 million workers.

UK SOC UK Occupation Matched US Occupation AI Exposure