AI Labour Impact Observatory

Method

How we measure — and where we fall short

This section publishes, in four blocks, the methodological honesty floor of the observatory: what we measure, how, what we DON'T measure, and what could be wrong. It's the section we most want you to read before citing the IILE-IA.

Editorial indexIILE-IA v0.2Experimental — v0.1

IILE-IA — Exponential Labour Impact Index of AI

IILE-IA combines six vectors into a single comparable reading. Its purpose is not to forecast job destruction: it is to flag where transformation pressure is high AND the capacity to absorb it is low.

IILE-IA coverage

IILE-IA coverage by dimension

How many verified claims already inform each dimension. The bars are NOT a score: they count linked claims.

1. ETechnical exposure

   Weight 0.20

   • 2 claims linked
   • Initial evidence

2. AReal adoption

   Weight 0.15

   • 0 claims linked
   • Editorial reading

3. TTask transformation

   Weight 0.20

   • 2 claims linked
   • Initial evidence

4. SSkills velocity

   Weight 0.15

   • 1 claims linked
   • Initial evidence

5. MLabour-market signal

   Weight 0.15

   • 2 claims linked
   • Initial evidence

6. BAdaptation gap

   Weight 0.15

   • 0 claims linked
   • Editorial reading

Index not calibrated yet.

From source to decision

How a figure travels from the official source down to the editorial decision. Each step is a filter: if one stage fails, the decision does not stand.

1. 01S — Skill velocity

   1. Verified source

      World Economic Forum — Future of Jobs Report 2025 www.weforum.org

   2. Traceable claim

      WEF: 39 % of key skills change by 2030.

   3. IILE dimension

      S — Skill velocity

   4. Editorial insight

      The minimum-employability bar moves faster than the traditional training cycle.

   5. Suggested decision

      Prioritise continuous reskilling before the wage premium becomes a barrier to entry.

2. 02E — Technical exposure

   1. Verified source

      IMF — Gen-AI: Artificial Intelligence and the Future of Work www.imf.org

   2. Traceable claim

      IMF: ~40 % of global employment is exposed to AI.

   3. IILE dimension

      E — Technical exposure

   4. Editorial insight

      Exposure measures potential impact, not automatic substitution.

   5. Suggested decision

      Do not confuse exposure with job destruction: some work will be augmented.

3. 03T — Task transformation

   1. Verified source

      ILO — Generative AI and Jobs: A Refined Global Index of Occupational Exposure webapps.ilo.org

   2. Traceable claim

      ILO: refined index built on 52,558 task-level datapoints.

   3. IILE dimension

      T — Task transformation

   4. Editorial insight

      The relevant unit for labour impact is the task, not the job.

   5. Suggested decision

      Analyse task portfolios inside each role before any headcount decision.

4. 04M — Labour-market signal

   1. Verified source

      Stanford HAI — AI Index 2025 Economy chapter hai.stanford.edu

   2. Traceable claim

      Stanford HAI: AI Index 2025 Economy chapter as priority source.

   3. IILE dimension

      M — Labour-market signal

   4. Editorial insight

      Operational signals (vacancies, wages) precede aggregate effects.

   5. Suggested decision

      Track AI-skills demand as a leading indicator instead of waiting for the observed datum.

The flow does not imply automatic decisions: every step still requires editorial reading.

Reading by audience

Four editorial readings of the dashboard, one per audience. These are not automatic recommendations: they are starting points each audience must reconcile with their context.

• Business and executives

  What to watch

  Real adoption (A·45) and task transformation (T·70): T is already high — tasks inside each role are migrating fast. A·45 says organisational adoption hasn't yet generalised in production.

  What to decide

  Map tasks, not jobs: redesign task portfolios before any headcount decision. The A→T gap is the productivity-capture window before competitors close it.

  What not to overinterpret

  T·70 measures task transformation in exposed occupations, not job destruction. The 22% disruption-by-2030 figure is an employer forecast (WEF), not a layoff projection.

• Professionals (cognitive roles)

  What to watch

  Skill velocity (S·72) and labour-market signal (M·42): S is the highest score in the observatory — skills shift fast. M·42 says that shift has NOT yet translated into aggregate employment disruption (Yale Budget Lab).

  What to decide

  Update your skills portfolio continuously, prioritising AI-complementary skills. Today's AI-skills wage premium (+28% per Lightcast) is an open window, not a permanent state.

  What not to overinterpret

  Technical exposure (E·65) is not substitution: many tasks are augmented, not eliminated (Anthropic Economic Index: ~49% of jobs already use AI in ≥1/4 of tasks).

• Education and talent

  What to watch

  Skill velocity (S·72) and adaptation gap (B·52): S sets the market's pace; B·52 says organisations are NOT yet mature enough to absorb it (McKinsey: only 1% consider themselves AI-deployment mature).

  What to decide

  Accelerate reskilling cycles so the curriculum does not lag the skills market. Curricula need shorter cycles than the usual annual review.

  What not to overinterpret

  The 39% key-skills change by 2030 is employer expectation (WEF), not a prescribed curriculum. And B·52 is editorial reading, not a direct measure of organisational capacity.

• Public policy / institutions

  What to watch

  Adaptation gap (B·52) and technical exposure (E·65): E·65 is the second-highest score — high task exposure, especially in advanced economies (IMF: ~60% of employment exposed in developed economies).

  What to decide

  Watch the gap B before exposure E: the vulnerable population is not the exposed, but those without a safety net to absorb the transformation. Prioritise safety nets and active labour policies.

  What not to overinterpret

  The 60% exposure in advanced economies (IMF) is methodological, not a count of lost jobs. ILO 2025 disaggregates by gender: 9.6% female employment vs 3.5% male in high-exposure occupations.

Each reading anchors the IILE-IA v0.2 scores (confidence: low, provisional editorial reading) to an actionable recommendation. Next review: 2026-06-15. No reading is a statistical prediction.

Other readings

IILE-IA v0.2 in the index ecosystem

Four external readings the observatory tracks, each with its headline, scope, and our editorial position vs IILE-IA. We don't compete with them — we use them as anchors.

• International Monetary Fund · imf.org

  IMF AI Preparedness Index (AIPI)

  ~60 %

  Scope

  174 economies · global · annual

  Measures

  Employment exposed to AI in advanced economies. Composite index 0-1 across 4 pillars: digital infrastructure, human capital + labour policies, innovation + economic integration, regulatory + ethical framework.

  Vs. IILE-IA

  AIPI is composite and country-comparable; IILE-IA is NOT by design — it is an editorial multi-signal reading with inline caveats. If you need one comparable number for Spain vs Germany, AIPI is the tool. If you want to know what changed this month and why it matters for your role, IILE-IA is the reading.

• Stanford HAI · hai.stanford.edu

  Stanford AI Index 2026 — Economy

  Dedicated chapter

  Scope

  Global · macro synthesis · annual

  Measures

  AI investment, enterprise adoption, productivity and labour-market signals. The gold standard in narrative + chart-per-claim. Hype-vs-evidence side-by-side by default.

  Vs. IILE-IA

  The AI Index is the broadest evidence corpus of the year, but lands as an annual PDF. IILE-IA picks its most relevant claims (Anthropic Economic Index, enterprise adoption) and re-anchors them monthly alongside the other Tier 1-3 sources.

• PwC · pwc.com

  PwC AI Jobs Barometer 2025

  +56 %

  Scope

  ~1B postings · 6 continents · annual

  Measures

  AI-skills wage premium in exposed occupations, from a billion-posting analysis. Also 66 % acceleration in skills change (vs non-AI-exposed).

  Vs. IILE-IA

  PwC is pure market signal (M in IILE-IA) and the best reference for the 'AI premium'. The Tier 4 bias (consultancy PR-packaging) is declared in the method; we use the number but triangulate it with the Lightcast Disruption Matrix.

• ILO · NASK · ilo.org

  ILO 2025-NASK Refined Index

  9.6 % vs 3.5 %

  Scope

  Global · occupational · by gender

  Measures

  Refined occupational GenAI exposure by task, disaggregated by gender and country. 9.6 % female employment vs 3.5 % male in high-exposure occupations.

  Vs. IILE-IA

  ILO 2025 is our canonical anchor for the E (technical exposure · 65) dimension of IILE-IA. Gender disaggregation is a methodology IILE-IA does NOT currently replicate — one of the gaps the Método section declares explicitly.

When a headline claims 'AI will destroy X jobs', cross-check against at least two of these four. If all four disagree with the headline, it's almost always overclaim.