Living dashboard
AI Labour Impact Observatory
AI's impact on employment
A living dashboard tracking the impact of artificial intelligence on employment, combining global sources, market signals and our own interpretation.
This observatory does not measure jobs created or destroyed alone. It measures the pressure of labour transformation and the capacity to absorb it.
- IILE-IA v0.2
- Curated monthly
- Weekly source watcher
- Experimental
Observatory evolutionLatest snapshot
Observatory evolution
Each meaningful update will produce an editorial snapshot: what changed, which source backs it, which indicator moves and how confident we are.
Y axis: 0–100. Editorial bands (Low pressure → Critical fracture) shaded in background. Each dot is a reviewed snapshot.
With a single numeric snapshot, the chart shows no trend yet. The monthly cadence will fill it in.
v0.2-editorial-provisional
- Methodology: IILE-IA v0.2
- Hand-curated
- Editorial provisional
- Taken on 2026-05-15
Total score
58.7
Components
E
65
A
45
T
70
S
72
M
42
B
52
Editorial baseline
- Hand-curated
- Pending calibration
- Taken on 2026-05-05
First editorial snapshot. No calibrated score until sources and indicators are verified.
Next milestone: first verified-claims slice that can initialise IILE-IA calibration.
What changed
- Manual entry
- Pending refresh
- Next version: real per-source signals
What changed. Why it matters. Whom it affects. What decision it suggests.
Monthly editorial cadence from 2026-05-15. The automated Tier 1–3 source watcher lands in PR 5 of the restructure.
2026-05-15
IILE-IA v0.2 — first editorial numeric reading per dimension.
Until today every dimension was `null`. v0.2 publishes six editorial numbers (E·65, A·45, T·70, S·72, M·42, B·52) with confidence: low and a total of 58.7. These are a provisional editorial reading, not a statistical calibration.
Observatory readers — the dashboard now shows a concrete editorial position per dimension, not just `pending`.
Treat the reading as a monthly-revisable editorial hypothesis, not a calibrated measurement. Real numeric calibration (Phase 6) needs ≥3 real snapshots.
2026-05-15
Dashboard visual restructure: 6 tabs → 4, compact hero, tabs in the first fold.
The first fold had 4 stacked blocks (hero + summary strip + calibration status + CTA) pushing the tab strip ~700px below the fold. The restructure moves the summary strip + calibration status INSIDE the Reading tab, compresses the hero, and leaves the tab strip visible on first scroll.
Full dashboard reading experience — starting to read no longer requires scrolling.
Promote the editorial Reading as the first impression and keep Evidence / Method / Evolution as opt-in depth.
2026-05-04
WEF Future of Jobs Report 2025 added to the observatory.
WEF projects 22% of jobs will change by 2030 (170M new roles, 92M displaced) and 59 in 100 workers will need reskilling/upskilling before 2030.
Professionals, HR, and corporate L&D.
Plan reskilling as ongoing cadence in the operating portfolio (SOX), not as an event.
2026-05-04
IMF: ~40% of global employment exposed to AI; ~60% in advanced economies.
Exposure is not destiny; some exposed jobs benefit from productivity, others face lower demand.
Public-policy design and corporate planning in advanced economies.
Distinguish exposure from observed outcomes when reading headlines.
2026-05-04
Anthropic Economic Index: ~49% of jobs saw at least 25% of tasks done with Claude (March 2026).
Platform telemetry — real adoption signal, not projection.
Knowledge work and professional services already using AI in production.
Move the observatory from projection-mode to sustained measurement.
2026-05-04
Yale Budget Lab + Brookings: aggregate disruption NOT yet visible in US labour market.
Empirical counterweight to the immediate-fracture narrative.
Editorial reading of the observatory; helps avoid overclaim.
Keep IILE-IA as an experimental editorial index, not a fracture forecast.
2026-05-04
PwC + LinkedIn + Lightcast: AI skills wage premium and skills-velocity already observable.
Aggregate disruption isn't visible yet, but value is shifting toward AI and hybrid skills.
Professionals with cognitive-routine profiles without AI upskilling.
Accelerate upskilling before the premium becomes a barrier to entry.
Source map
Sources are ordered by methodological strength. The editorial update cadence respects that order: what enters the index is not the same as what triggers an alert.
Tiers 1–3 calibrate the IILE-IA index. Tier 4 helps interpret. Tier 5 adds adoption signals. Tier 6 triggers alerts, but does NOT recalculate the index.
- Tier 1Calibrates the index
Multilaterals and official statistics
Bodies with public responsibility for labour and economic data.
Sources
- WEF
- ILO
- OECD
- IMF
- World Bank
- Eurostat
- BLS
- O*NET
Tier 1–3 can influence calibration of the IILE-IA index.
- Tier 2Calibrates the index
Labour-market platforms
Operational employment signals: vacancies, salaries, skills dynamics.
Sources
- Indeed
- Lightcast
- ADP
- Revelio
- Randstad Research
Tier 2 contributes high-frequency operational signals; useful for the Labour-market section.
- Tier 3Calibrates the index
Academic research and think tanks
Peer-reviewed work and reference research centres.
Sources
- Stanford HAI
- NBER
- MIT
- MIT Iceberg
- Yale Budget Lab
- Brookings
- Funcas
Tier 3 contributes conceptual frameworks and empirical validation for interpretation.
- Tier 4Helps interpret
Consulting and enterprise research
Enterprise surveys and internal-adoption models; useful but with client bias.
Sources
- PwC
- McKinsey
- BCG
- Deloitte
- Accenture
- Microsoft
Tier 4 can influence interpretation but is weighted carefully due to client bias.
- Tier 5Adds adoption signals
AI platform telemetry
First-party usage data from those who provide the models and tools.
Sources
- Anthropic Economic Index
- OpenAI
- GitHub
- Microsoft Copilot
Tier 5 can flag adoption and productivity patterns without entering official statistics.
- Tier 6Triggers alerts; does not recalibrate
Media and curated analysis
Early detectors: they amplify signals rather than generating them.
Sources
- FT
- The Economist
- NYT
- HBR
- Newsletters
Tier 6 can trigger editorial alerts and signals but does NOT update the index directly.
Source coverage by tier
How many verified sources we hold per tier and the role each tier plays in the IILE-IA index.
Tier 1Calibrates the index
Sources
15
With canonical link
7
Calibrates the IILE-IA index
Tier 2Calibrates the index
Sources
6
With canonical link
1
Calibrates the index (market operational signals)
Tier 3Calibrates the index
Sources
10
With canonical link
5
Calibrates the index (academic + empirical frameworks)
Tier 4Helps interpret
Sources
6
With canonical link
0
Helps interpret (consulting / client bias)
Tier 5Provides adoption signals
Sources
6
With canonical link
1
Adds adoption signals (telemetry)
Tier 6Triggers alerts (does NOT recalibrate)
Sources
5
With canonical link
0
Triggers alerts; does NOT recalibrate the index
Source registry
Editorial directory of the sources the observatory tracks. Each source declares its type, what it measures best, its cadence, and verification status.
Verified: link and metric reviewed. Official source linked: canonical URL confirmed; concrete metrics still pending claim extraction. Curated from operator brief: editorial proposal pending URL + date verification before quoting as definitive. Pending verification: needs review before being treated as final.
Tier 1Calibrates the IILE-IA index
World Economic Forum
- Curated from brief
- High confidence
- Type
- Multilateral / global reports
- Best for
- Future of Jobs Report — macro projections
- Why it matters
- Macro reference for job creation/displacement, skills gap and reskilling.
- Cadence
- Annual + thematic updates
ILO
- Curated from brief
- High confidence
- Type
- Multilateral / official labour statistics
- Best for
- Occupational exposure to GenAI by task
- Why it matters
- Combines task-level data, expert judgement and AI estimates with a decent-work lens.
- Cadence
- Annual / by topic
OECD
- Curated from brief
- High confidence
- Type
- Multilateral / public policy
- Best for
- Changes in tasks, skills and policy
- Why it matters
- Fine-grained view of how work content changes in exposed occupations.
- Cadence
- Annual / per study
IMF
- Curated from brief
- High confidence
- Type
- Multilateral / economic analysis
- Best for
- Macro exposure and inequality by country type
- Why it matters
- Distinguishes advanced vs. emerging economies in exposure and productivity effects.
- Cadence
- Annual / staff papers
World Bank
- Pending verification
- Medium confidence
- Type
- Multilateral / development
- Best for
- Employment, productivity and skills in emerging economies
- Why it matters
- Complements the IMF with a focus on developing countries and employment policy.
- Cadence
- Annual / per report
Eurostat
- Pending verification
- High confidence
- Type
- EU official statistics
- Best for
- Official EU employment and skills
- Why it matters
- Quantitative anchor for the observatory's European reading.
- Cadence
- Quarterly / annual
BLS
- Pending verification
- High confidence
- Type
- US official statistics
- Best for
- Employment, wages, productivity and occupations (US)
- Why it matters
- US quantitative anchor — occupational projections and BLS Employment Projections.
- Cadence
- Monthly / biennial
O*NET
- Pending verification
- High confidence
- Type
- US occupational taxonomy
- Best for
- Tasks, skills and attributes by occupation
- Why it matters
- Structural base on which almost every AI-exposure paper builds its analysis.
- Cadence
- Quarterly
World Economic Forum — Future of Jobs Report 2025
- Official source linked
- Medium confidence
- Type
- Global employer survey
- Best for
- Jobs and skills outlook, sector trends and employer expectations
- Why it matters
- Macro reference for job creation/displacement, skills gap and reskilling.
- Cadence
- Biennial
- Dashboard use
- Executive radar
- Source radar
- IILE component: Skills
- Use for macro trend context and skills/employment transformation outlook.
ILO — Generative AI and Jobs: A Refined Global Index of Occupational Exposure
- Official source linked
- Medium confidence
- Type
- Occupational exposure research
- Best for
- Occupational exposure, task-level analysis, automation vs. augmentation framing
- Why it matters
- Provides a verified methodology for generative-AI exposure at the task level.
- Cadence
- Per publication
- Dashboard use
- Evidence vs. hype
- IILE component: Exposure
- Use for exposure and task-based methodology.
IMF — Gen-AI: Artificial Intelligence and the Future of Work
- Official source linked
- Medium confidence
- Type
- Macroeconomic + labour analysis
- Best for
- Macro labour-market exposure, inequality risk, policy framing
- Why it matters
- Distinguishes advanced vs. emerging economies in exposure and productivity effects.
- Cadence
- Per staff note
- Dashboard use
- Evidence vs. hype
- IILE component: Adaptation gap
- Staff discussion note; present as policy/economic analysis, not as official VTE index calibration yet.
OECD — AI and work
- Official source linked
- Medium confidence
- Type
- Policy + public research
- Best for
- Job quality, workplace policy, adoption and governance
- Why it matters
- Provides institutional reading of how work content and policy responses change.
- Cadence
- Continuous
- Dashboard use
- Source radar
- IILE component: Adoption
- Use for policy and institutional readiness.
OECD — Future of work
- Official source linked
- Medium confidence
- Type
- Policy + public research
- Best for
- Future of work context, skills policy, job transition framing
- Why it matters
- OECD editorial hub on work transformation and public policy.
- Cadence
- Continuous
- Dashboard use
- Executive radar
- Source radar
- Use as broader context source.
ILO — Generative AI and Jobs (2025 update, with NASK)
- Official source linked
- High confidence
- Type
- Occupational exposure research (refined method)
- Best for
- Refined occupational exposure to GenAI, disaggregated by gender and country
- Why it matters
- 9.6% female employment vs 3.5% male employment in high-exposure occupations; 'transformation, not destruction' framing anchored to national micro-data.
- Cadence
- Per update (≥annual)
- Dashboard use
- Evidence vs. hype
- IILE component: Exposure
- IILE component: Task transformation
- Gender disaggregation is robust — use for equity caveats on every relevant KPI.
Eurostat — AI use by enterprise size
- Official source linked
- High confidence
- Type
- EU official statistics
- Best for
- EU enterprise AI adoption by size — official figure
- Why it matters
- Official figure: ~20% of EU firms use AI in 2025 (vs ~13% in 2024). Single public quantitative anchor for the European adoption reading.
- Cadence
- Annual
- Dashboard use
- Global pulse
- IILE component: Adoption
- Public API, no paywall. Suitable for auto-refresh in PR 5 (when the source-watcher lands).
Tier 2Calibrates the index (market operational signals)
LinkedIn Economic Graph
- Curated from brief
- Medium confidence
- Type
- Labour platform / professional graph
- Best for
- Skills, career paths, mobility and emerging occupations
- Why it matters
- Estimates that 70% of the skills used in most jobs will have changed by 2030.
- Cadence
- Continuous / annual reports
Indeed Hiring Lab
- Curated from brief
- Medium confidence
- Type
- Labour platform / job postings
- Best for
- Skills transformation in job postings
- Why it matters
- GenAI Skill Transformation Index: 26% of jobs could be highly transformed by GenAI; 46% of skills sit in the hybrid zone.
- Cadence
- Continuous / quarterly reports
Lightcast
- Curated from brief
- Medium confidence
- Type
- Labour platform / skills intelligence
- Best for
- Skills demand in job postings + wage premium
- Why it matters
- Across 1,300M postings: 28% wage premium for AI skills (~$18K/year); 51% of AI postings are outside IT.
- Cadence
- Continuous
ADP
- Curated from brief
- Medium confidence
- Type
- Payroll platform / labour market
- Best for
- Employment, wages and turnover from real payroll data
- Why it matters
- Engagement, stress and worker perception among AI users vs. non-users.
- Cadence
- Monthly / reports
Revelio
- Pending verification
- Medium confidence
- Type
- Labour platform / workforce analytics
- Best for
- Talent movements, turnover and workforce composition
- Why it matters
- Alternative view to Hiring Lab and LinkedIn for mobility trends.
- Cadence
- Continuous
Randstad Research — AI and the labour market (Spain)
- Official source linked
- Medium confidence
- Type
- Talent-market research (Spain)
- Best for
- Skills demand, wages and turnover with ES talent focus
- Why it matters
- Employer-employee perspective complementing Funcas, with ES-specific talent angle.
- Cadence
- Continuous / annual reports
- Dashboard use
- IILE component: Skills
- IILE component: Market signal
- Useful for cross-checking Randstad vs Funcas signals and spotting discrepancies in the ES reading.
Tier 3Calibrates the index (academic + empirical frameworks)
▸ Show Tier 3 sources (10)
Stanford HAI
- Curated from brief
- High confidence
- Type
- Academic research / AI Index
- Best for
- Population adoption and early labour-market evidence
- Why it matters
- GenAI reached 53% population adoption in 3 years (faster than the PC or the internet); evidence on young workers in exposed occupations.
- Cadence
- Annual + papers
NBER
- Pending verification
- High confidence
- Type
- Peer-reviewed economic research
- Best for
- Working papers on AI and the labour market
- Why it matters
- Conceptual frameworks and empirical validation for interpretation.
- Cadence
- Continuous / per paper
MIT
- Pending verification
- High confidence
- Type
- Academic research / AI-and-work centres
- Best for
- Worker productivity, occupational exposure
- Why it matters
- Influential frameworks for measuring AI impact on tasks and productivity.
- Cadence
- Continuous / papers
Yale Budget Lab
- Curated from brief
- High confidence
- Type
- Economic policy research
- Best for
- Empirical counterweight to the hype: what is actually observed in the market
- Why it matters
- Finds no substantial acceleration in the composition of the US labour market since ChatGPT.
- Cadence
- Continuous / per analysis
Brookings
- Curated from brief
- High confidence
- Type
- Public policy research
- Best for
- Editorial synthesis on AI and employment
- Why it matters
- Stresses that the aggregate evidence remains inconclusive and that better data is needed.
- Cadence
- Continuous
Stanford HAI — AI Index 2025 Economy chapter
- Official source linked
- Medium confidence
- Type
- Labour-market data analysis
- Best for
- AI labour demand, job postings, skills signals
- Why it matters
- Dedicated economy and labour-market chapter of the AI Index 2025.
- Cadence
- Annual
- Dashboard use
- Global pulse
- IILE component: Market signal
- Use for labour-demand signals; exact figures require later extraction and validation.
Funcas — AI and labour market in Spain
- Official source linked
- Medium confidence
- Type
- Applied research (Spain)
- Best for
- Occupational exposure, jobs destroyed vs created, enterprise adoption in Spain
- Why it matters
- Central scenario: −1.7M to −2.3M jobs destroyed and +1.61M created in Spain 2025-2035 (net −400K). 21.1% of firms ≥10 employees use AI in Q1 2025 (vs 12.4% in 2023).
- Cadence
- Annual / per study
- Dashboard use
- Global pulse
- IILE component: Market signal
- IILE component: Adoption
- Critical anchor for ES readings. Always publish gross-destruction + gross-creation + net, never just net.
Stanford AI Index 2026 — Economy chapter
- Official source linked
- High confidence
- Type
- Academic review
- Best for
- Macro synthesis of adoption, investment, productivity and employment
- Why it matters
- Gold standard for narrative + chart-per-claim in AI economics. Hype-vs-evidence side-by-side by default.
- Cadence
- Annual
- Dashboard use
- Global pulse
- Evidence vs. hype
- IILE component: Adoption
- Compare against v2025 (chapter4_final.pdf) to detect year-over-year deltas in the macro reading.
MIT FutureTech — Project Iceberg
- Official source linked
- Medium confidence
- Type
- Academic simulation
- Best for
- Wage-value-at-risk by skill × occupation × county (US)
- Why it matters
- Large Population Models simulation: 151M workers × 32K skills × 3K counties. Iceberg metaphor (automatable skills below waterline) for visualising magnitude without binary headlines.
- Cadence
- Per publication / update
- Dashboard use
- Evidence vs. hype
- IILE component: Skills
- IILE component: Market signal
- Iconic image useful for outreach. US-only — extrapolate carefully to other geographies.
- Official link
- https://iceberg.mit.edu/report.pdf
Yale Budget Lab — Tracking the Impact of AI on the Labor Market
- Official source linked
- High confidence
- Type
- Policy-economics research
- Best for
- Empirical monthly counterweight: is disruption actually visible in the US labour market?
- Why it matters
- Monthly US CPS series measuring occupational / industry dissimilarity and exposure. Recent verdict: AI has NOT yet dented aggregate employment.
- Cadence
- Monthly
- Dashboard use
- Evidence vs. hype
- IILE component: Market signal
- Key editorial calibrator against 'inevitable fracture' narratives. They also publish 'what we know / what we don't' as a formal artifact.
Tier 4Helps interpret (consulting / client bias)
▸ Show Tier 4 sources (6)
PwC
- Curated from brief
- Medium confidence
- Type
- Consulting / Global AI Jobs Barometer
- Best for
- Postings, wages and skills change in exposed jobs
- Why it matters
- ~1,000M postings analysed; 62% wage premium for AI skills and 116% acceleration in skills change.
- Cadence
- Annual
McKinsey
- Curated from brief
- Medium confidence
- Type
- Consulting / Superagency
- Best for
- Organisational maturity and governance
- Why it matters
- Almost every company is investing in AI; only 1% consider themselves mature in deployment.
- Cadence
- Annual / thematic reports
BCG
- Pending verification
- Low confidence
- Type
- Consulting
- Best for
- Value captured and change management
- Why it matters
- Corporate reading complementary to McKinsey/PwC.
- Cadence
- Annual
Deloitte
- Pending verification
- Low confidence
- Type
- Consulting
- Best for
- Surveys of leaders and of the workforce
- Why it matters
- Additional sector / regional reading.
- Cadence
- Annual
Accenture
- Pending verification
- Low confidence
- Type
- Consulting / Technology Vision
- Best for
- Enterprise technology adoption
- Why it matters
- Cross-reading with Microsoft/PwC surveys.
- Cadence
- Annual
Microsoft
- Curated from brief
- Medium confidence
- Type
- Technology company / Work Trend Index
- Best for
- Corporate adoption, agents and new models of work
- Why it matters
- 33% of leaders are considering headcount cuts, 78% are considering hiring for new AI roles, 83% believe AI will let people take on more complex work earlier in their careers.
- Cadence
- Annual
Tier 5Adds adoption signals (telemetry)
▸ Show Tier 5 sources (6)
Anthropic
- Curated from brief
- Medium confidence
- Type
- AI platform telemetry
- Best for
- Real AI use by task and occupation (Anthropic Economic Index)
- Why it matters
- ~49% of jobs have seen at least 1/4 of their tasks performed with Claude (March 2026).
- Cadence
- Continuous
OpenAI
- Pending verification
- Medium confidence
- Type
- AI platform telemetry
- Best for
- Real use, declared occupational exposure
- Why it matters
- Co-publishes exposure studies with researchers; platform usage signals.
- Cadence
- Continuous
GitHub
- Pending verification
- Medium confidence
- Type
- Developer telemetry / Copilot
- Best for
- Productivity and adoption in software development
- Why it matters
- Direct usage data from developers using Copilot.
- Cadence
- Continuous
Microsoft Copilot
- Pending verification
- Medium confidence
- Type
- Productivity platform telemetry
- Best for
- Corporate adoption of copilots in M365
- Why it matters
- Usage metrics across office workflows.
- Cadence
- Continuous
Google
- Pending verification
- Low confidence
- Type
- Telemetry / Workspace AI
- Best for
- Adoption in Workspace and assisted search
- Why it matters
- Cross-reading with Microsoft Copilot.
- Cadence
- Continuous
Anthropic Economic Index
- Official source linked
- Medium confidence
- Type
- Platform telemetry (verified)
- Best for
- Real AI usage (Claude) by tasks mapped to O*NET
- Why it matters
- ~49% of occupations see at least 25% of tasks done with Claude (March 2026). First-party telemetry mapped to O*NET taxonomy — distinguishes task-level usage from occupation-level exposure.
- Cadence
- Continuous
- Dashboard use
- IILE component: Task transformation
- IILE component: Adoption
- Global pulse
- Tier 5 (private platform telemetry) — high adoption signal, lower methodological transparency vs ILO. Triangulate with LinkedIn + Lightcast.
- Official link
- https://www.anthropic.com/economic-index
Tier 6Triggers alerts; does NOT recalibrate the index
▸ Show Tier 6 sources (5)
Financial Times
- Pending verification
- Medium confidence
- Type
- Specialist press
- Best for
- Business and markets coverage
- Why it matters
- Early detector of corporate signals.
- Cadence
- Daily
The Economist
- Pending verification
- Medium confidence
- Type
- Specialist press
- Best for
- Macro editorial synthesis
- Why it matters
- Early detector of editorial framings and dominant narratives.
- Cadence
- Weekly
New York Times
- Pending verification
- Low confidence
- Type
- General press
- Best for
- Social and political coverage of AI impact
- Why it matters
- Captures the public and opinion angle.
- Cadence
- Daily
HBR
- Pending verification
- Low confidence
- Type
- Editorial press / management
- Best for
- Cases and management frameworks
- Why it matters
- Useful for extracting adoption and leadership frameworks.
- Cadence
- Weekly / monthly
Curated newsletters
- Pending verification
- Low confidence
- Type
- Curated / expert analysis
- Best for
- Fast detector of shifts and editorial nuance
- Why it matters
- Generates alerts; does not calibrate the index.
- Cadence
- Weekly
These sources are verified as a starting point. Concrete figures will be incorporated only when the claim, the date and the methodology are traced.
Reading guide
How to read this dashboard
What it is for
To recognise the pressure of labour transformation by sector, occupation and profile — and the capacity to absorb it. A curated editorial reading, not a prediction.
What it is NOT for
It does not predict job destruction. It is not an official statistical index. It does not replace Tier 1–3 sources; it weights them with explicit editorial weights.
How to use it while reading the book
Come back here whenever a chapter cites a specific vector (exposure, adoption, gap) to situate it in the wider picture and compare the rhythm across sectors.
How it will evolve
Future iterations will add real per-source signals, a declared editorial cadence, and an automated watcher for Tier 1–3. IILE-IA calibration will remain editorial.