Call Us: 020 4600 5848
Get Instant Quote

What this study does

We stitch together official UK employment by occupation, extend it to 2035 using the government’s Skills Imperative baseline, then overlay credible AI impacts using published task-exposure indices and two adoption paths (a conservative WEF-aligned curve and a faster, McKinsey-aligned curve). Finally, we redistribute adoption by occupation using an explicit weight so the same economy-wide adoption bites harder in clerical/admin roles than in skilled manual work. The occupational weighting is informed by McKinsey’s evidence that generative-AI adoption and value concentrate in customer operations and back-office functions, and are lower in physical/manual trades. McKinsey & Company

1) Data sources

  • Employment by occupation (historical, SOC 2020, with male/female split)
    Annual Population Survey (APS) via NOMIS. We use “Employment by occupation (SOC2020) by sex” (APS218) as the baseline for 2010–2025. Nomis Web

  • Baseline projections to 2035 (no-AI “business-as-usual”)
    UK Skills Imperative 2035 occupational projections (rolled to SOC 2-digit). We use the official series and programme documentation as the no-automation counterfactual for 2026–2035. GOV.UK+1

  • Task-level AI exposure (what could be automated in principle)
    ILO working paper / refined global index of occupational exposure to generative AI at ISCO-08, cross-walked to SOC 2020. International Labour Organization+1

  • Yearly adoption paths (what employers actually deploy)

    • WEF 2023 adoption trajectory for the conservative scenario. World Economic Forum+1

    • McKinsey 2023 accelerated scenario for the high-adoption case; we translate their “share of work time technically automatable” into an adoption path of automatable tasks (A) that reaches ≈0.85 by 2035 so that Exposure×Adoption ≈ 60% of total hours. McKinsey & Company+1

  • Classification mapping (ISCO ↔ SOC)
    ONS guidance on mapping SOC 2020 to ISCO-08 informs the crosswalk aligning exposure indices with UK SOC. Office for National Statistics

We reviewed recent IMF/OECD research on AI exposure and displacement to set reasonable displacement elasticities (δ) used in sensitivity tests. IMF

 

2) Preparing the baseline (2010–2035)

  1. 2010–2025 (historical): Aggregate APS employment to SOC 2-digit for All persons, Male, Female. Harmonise to SOC 2020 across the time series (rebasing early years where needed). Nomis Web

  2. 2026–2035 (no-AI baseline): Append Skills Imperative 2035 projections to create the business-as-usual counterfactual that already embeds population, sectoral mix and long-run occupational trends, but does not include AI displacement. GOV.UK

3) How AI impact is applied

Core equation (non-compounding, year by year)

For each occupation o and year y:

Employmento,y=Baselineo,y×(1−δ×Exposureo×Ay×wocc,o)\text{Employment}_{o,y}=\text{Baseline}_{o,y}\times\big(1-\delta\times \text{Exposure}_{o}\times A_{y}\times w_{\text{occ},o}\big)

  • Baselineo,y_{o,y}: Skills Imperative value for year y. GOV.UK

  • Exposureo_o: ISCO-based exposure mapped to SOC 2020; the share of tasks automatable in principle. International Labour Organization+1

  • AyA_y (adoption): economy-wide share of automatable tasks actually automated in year y.

    • Conservative (WEF): 2026→2035 ramp (e.g., 0.12 → 0.56). World Economic Forum

    • High-adoption (McKinsey-aligned): accelerated ramp (e.g., 0.12 → ~0.85 by 2035). We calibrate so Exposure×Adoption ≈ 0.60 in 2035, consistent with McKinsey’s estimate that current technologies (incl. GenAI) can automate ~60% of work time; this avoids double-discounting. McKinsey & Company

  • woccw_{\text{occ}} (adoption weight by occupation): redistributes impact, informed by McKinsey’s evidence that adoption/value are concentrated in customer operations and back-office functions.

    • Clerical/admin & customer service: 1.3
    • Skilled trades, production & drivers: 0.8
    • All other groups: 1.0
      (Optionally normalised so the employment-weighted mean ≈ 1 per year.) McKinsey & Company

  • δ\delta (task→job displacement elasticity): share of automated tasks translating to job losses rather than redeployment. We use 0.7 in the conservative run and 0.8 in the high-adoption run (upper-bound in recent literature). IMF

Non-compounding choice. We deliberately do not compound reductions; each year’s adjustment is applied to that year’s Skills Imperative baseline. This keeps the model anchored to the official projection rather than an internally snowballing series.

4) Scenarios we publish

  • Conservative (WEF-aligned)
    Adoption AyA_y: 0.12 → 0.56 (2026–2035); δ=0.7\delta=0.7; woccw_{\text{occ}} as above. Policy-aligned, moderate adoption pace. World Economic Forum

  • Accelerated / high-adoption (McKinsey-aligned)
    Adoption AyA_y: 0.12 → ≈0.85 (2026–2035); δ=0.8\delta=0.8; woccw_{\text{occ}} as above. Calibrated so Exposure×Adoption ≈ 60% of total hours automated by 2035. McKinsey & Company

(All headline statistics clearly state which scenario they come from; both tables are downloadable below for transparency.)

5) Gender results (how we compute them)

For each scenario and year we publish All persons, Male, Female totals by SOC 2-digit and as whole-economy sums. Headline % changes are computed 2025 → 2035.

6) Quality checks we apply (and publish)

  • Back-test: with Ay=0A_y=0 the model reproduces the Skills Imperative baseline exactly. GOV.UK

  • Aggregate calibration (2035, accelerated): compute the employment-share-weighted Exposure×Adoption; target ≈0.60 of hours automated, per McKinsey. McKinsey & Company

  • Accounting checks: Male + Female = All persons by SOC and year; the weight normaliser (if used) keeps system-wide adoption unchanged when redistributing across occupations.

7) Limitations & interpretation notes

  • Exposure indices are estimates and may be revised; we do not alter them. International Labour Organization
  • Skills Imperative is a trend baseline; macro shocks or policy changes could move the counterfactual. GOV.UK
  • We model task displacement, not new task creation or general-equilibrium feedbacks; results are an impact overlay.
  • No regional splits (London vs UK) and no firm-level adoption heterogeneity in this release.

8) Reproducibility & downloads

 

9) Why we commissioned this work

As a London electrical firm investing in apprenticeships and technical training, Electricians London 247 commissioned this analysis to understand which sectors and skill levels will define the next generation of UK work. The study was developed in collaboration with independent data analysts and economists to ensure the modelling reflects credible, evidence-based assumptions.