The Skills Divide: What the IMF, ILO, and OECD's Joint Workforce Analysis Says About AI's Real Workforce Impact

Executive Summary

• 1 in 10 job vacancies in advanced economies now requires at least one AI-era skill — a figure that was nearly zero before 2015. In emerging markets, the rate is roughly half that. The gap between countries that can supply these skills and those that cannot is widening fast.
• AI-developer skills carry a 7.5–8% wage premium in the UK and above 8% in the US. AI-user skills carry a 2% premium in the US and 7.5–8% in the UK. Workers with multiple new skills reach premiums of 15.1% (UK) and 8.5% (US). The skills premium is real, measurable, and compounding.
• AI skills do not increase overall employment — they redistribute it. Occupations with high AI exposure and low human-AI complementarity see 3.6% lower employment over five years. Middle-skilled white-collar workers and young workers bear the brunt.
• The US and Denmark lead global AI skill adoption; the US is the origination hub. UK and Denmark adopt within 2–4 months of US postings. Germany lags further. Brazil and South Africa lag 8–9 months. Within the US, California, New York, and Texas are the skill generation centers.
• Generative AI is the #1 fastest-growing skill in every advanced economy tracked. It topped the 2024 job-posting growth charts in the US, UK, and all other measured countries — the first time a single technology skill has unified the global growth list.

What the Study Is and Why It Matters

The IMF Staff Discussion Note SDN/2026/001 is a tri-organization collaboration between the International Monetary Fund, the International Labour Organization, and the OECD — the three bodies that collectively set the standard for cross-country labor market analysis. Published January 9, 2026, it is the most comprehensive government-sourced analysis of AI skill demand and supply dynamics available.

The methodology is not a survey of executive opinions. It is a direct analysis of Lightcast job postings from 220,000+ websites globally — covering the US from 2010 to 2025, the UK from 2012 to 2024, and Germany, Brazil, Denmark, and South Africa from 2021 to 2024. That data is cross-validated against American Community Survey records, German administrative labor records, ILO data, OECD skill surveys, and 18 million Lightcast worker profiles. The authors then run wage regressions, event studies, and employment effect models to isolate what AI skill demand actually does to wages and headcount.

The credibility of this source is the highest available in the corpus: three independent government organizations, no commercial sponsor, primary job-posting data rather than self-reported surveys, and peer-reviewed methodology. For CHROs and CFOs who need a non-vendor, non-consulting anchor for workforce investment decisions, this is that anchor.

The AI Skills Premium: Real Numbers for Real Budget Decisions

The wage premium data is the most immediately actionable finding for a mid-market CHRO building an AI upskilling business case.

AI-developer skills — defined as skills to build, train, and deploy AI systems — carry premiums of 7.5–8% in the UK and above 8% in the US. At a median software engineer salary of $130,000, that is a $10,400–$10,800 per year premium. At a mid-market scale of 10 engineers, that is a $100,000+ annual compensation gap against any firm that cannot develop or retain AI-capable engineers.

AI-user skills — defined as the ability to deploy AI tools in non-AI-specific work — carry a 2% premium in the US and 7.5–8% in the UK. The US-UK gap is notable: AI-user skills are more common in the US market, so the premium has compressed. In the UK, where supply is thinner, the same skills earn nearly four times the premium. US companies with UK or EU operations face a significantly higher AI-user skill cost.

Workers with multiple new skills earn premiums of 15.1% in the UK and 8.5% in the US. This is the compounding premium: acquiring a second AI-relevant skill does not add 2% + 2%, it multiplies.

The practical implication for a 500-person mid-market company: upskilling your existing workforce before the premium compresses further is cheaper than waiting and hiring at market rate. The ATD benchmark puts AI training cost at $165 per learning hour; a 12-hour curriculum costs $1,980 per employee. The 2% wage premium at a $70,000 median salary is $1,400 per year — the ROI is marginal. But the 8%+ developer premium at $130,000 is $10,400 per year, making training investment pay back inside six months for any engineer who stays.

The Employment Effect: What “AI Doesn’t Eliminate Jobs” Actually Means

The IMF study finds that greater AI skill demand does not increase overall employment. This is distinct from what happens with other new skills, which do increase employment (a 1-percentage-point increase in new skill postings is linked to a 1.3% employment gain in the US).

What AI skill demand does instead is redistribute employment. Occupations with high AI exposure and low human-AI complementarity — meaning occupations where AI can substitute for rather than augment the work — show 3.6% lower employment after five years. That 3.6% figure is an average across the affected occupational cluster, not an outlier.

The occupations most exposed are white-collar middle-skilled roles: the paralegal reviewing standard contracts, the financial analyst compiling variance reports, the administrative coordinator managing scheduling and document routing. These are not routine factory tasks. They are the roles that mid-market companies hired entry-level and junior workers to perform.

Two downstream effects warrant attention:

Entry-level pipeline compression. The study finds AI skills are concentrated in workers with bachelor’s degrees (85%) and heavily concentrated in ICT/STEM graduates (60% of AI-developer skill holders). AI-user skills are distributed more broadly but still skew toward educated workers. The pipeline for developing AI-capable talent from within a mid-market company depends on which roles you hire at entry level — and the entry-level hiring market for AI-adjacent roles is contracting, as the Brent Orrell data in the workforce planning wiki separately documents.

Youth employment pressure. The study specifically flags youth workers as bearing disproportionate risk from AI skill demand. Young workers are overrepresented in the middle-skilled white-collar roles facing displacement, and underrepresented in the AI-developer skill holder pool. This matters for mid-market companies running internship programs or building talent pipelines from recent graduates: the roles those graduates are trained to enter are changing faster than the curricula preparing them.

The Global Skill Diffusion Map: Where the US Stands

The US is the origination hub for AI-era skill demand. Within the US, California, New York, and Texas generate the skills that then diffuse to other states within months (down from 1–2 years prior to 2020). Generative AI topped the US job posting growth list in 2024; it was the #1 fastest-growing skill in the UK, Germany, Brazil, and South Africa as well.

The speed of diffusion has accelerated. Skills that took 1–2 years to propagate from California to other US states in the early 2010s now propagate in months. Internationally, advanced economies (Denmark, UK) adopt within 2–4 months of US. Germany lags further. Emerging markets lag 8–9 months.

The practical implication: AI skills that are premium-priced today will become standard requirements within 12–24 months. The window to invest in upskilling before market wages catch up is measured in months, not years. The study’s cross-country diffusion model has consistently predicted US skill demand being a leading indicator for European and global demand — AI skills are following the same pattern.

The 2024 US top-10 fastest-growing skills list (from the Lightcast data underlying the study): Generative AI, Google Cloud Platform, Site Reliability Engineering, Amazon Web Services, Power BI, CI/CD, Kubernetes — and non-AI skills (TikTok management, social media management, Express.js) that reflect adjacent workflow transformation. The AI/cloud stack dominates the technical half of the list; marketing/content stack dominates the non-technical half.

The Skill Readiness Gap: Why Supply Doesn’t Match Demand

The study constructs a Skill Readiness Index across 24 countries measuring three dimensions: graduate output in relevant fields, adult literacy and numeracy, and availability of retraining infrastructure. Ireland, Finland, and Denmark lead. Chile, Italy, and Hungary trail.

The US sits in the moderate tier on this index despite leading on AI skill demand — a structural tension. The US generates the highest demand and originates the most AI skills, but has weaker retraining infrastructure and lower adult STEM literacy than Scandinavian peers. The implication: US companies cannot rely on the external labor market to supply AI-ready talent at the rate demand is accelerating. Internal upskilling is not optional — it is the only supply channel that scales at mid-market.

The Skill Imbalance Index classifies countries by whether their demand exceeds their supply or vice versa:

• High demand, low supply (action needed: expand training): Sweden, Luxembourg, Netherlands. US edges toward this quadrant as AI demand accelerates faster than graduate output.
• High supply, low demand (action needed: stimulate innovation): Ireland, Poland, Australia.
• Balanced: Chile, Estonia.

For US-based mid-market CHROs, the takeaway is structural: the external market will not close the skills gap at your company’s required pace. The policy prescription — lifelong learning, reskilling, labor mobility — maps to what the most successful mid-market deployers are already doing: DataCamp/YouGov’s 2x ROI multiplier from mature internal training programs reflects the same dynamic the IMF study identifies at the national level.

Key Data Points

Source credibility: HIGH. IMF/ILO/OECD tri-organization, government-published, no commercial sponsor. Primary job-posting analysis via Lightcast (220,000+ websites). Cross-validated against administrative records and population surveys. Methodology disclosed in full.

What This Means for Your Organization

Three decisions follow directly from this data.

Upskill before the premium compresses. The 2% AI-user wage premium in the US is evidence that supply is catching up to demand — but the 8%+ developer premium is evidence that the highest-value AI skills are still scarce. The window to train existing staff into AI-augmented roles at current market rates is closing. Companies that build internal capability now will carry lower fully-loaded talent costs than those that wait and hire at future market wage premiums.

Build your middle-skill transition plan. The 3.6% employment decline in high-exposure, low-complementarity occupations is not a decade away — the event study covers a five-year horizon from the point of AI skill adoption. For a mid-market company that uses entry-level paralegals, junior financial analysts, or administrative coordinators to handle high-volume routine work, the transition plan has to begin now. The question is not whether these roles change — it is whether the transition is managed deliberately or discovered after the fact.

Use the Skill Imbalance Index to frame the upskilling business case. The IMF’s finding that countries with high AI demand and low supply face the sharpest skills gaps maps directly to the company level: if your AI ambition exceeds your current employee skill base, the external labor market will not close the gap at speed. The DataCamp/YouGov data (n=517, February 2026) already in this corpus shows mature internal training programs achieve 2x the ROI rate of organizations relying on subscription-access training. The IMF study provides the macro-economic rationale for why that multiplier exists: supply is constrained, so internal production of skill is structurally advantaged.

If questions about how this connects to your specific workforce planning decisions — which roles to prioritize, what the training investment model looks like at your headcount — are worth a conversation, reach out directly at brandon@brandonsneider.com.

Sources

Primary Source

• IMF Staff Discussion Note SDN/2026/001, “Bridging Skill Gaps for the Future: New Jobs Creation in the AI Age,” Florence Jaumotte et al. (IMF, ILO, OECD collaboration), January 9, 2026. URL: https://www.imf.org/en/publications/staff-discussion-notes/issues/2026/01/09/bridging-skill-gaps-for-the-future-new-jobs-creation-in-the-ai-age-572136 Credibility: HIGH — tri-government-organization, no commercial sponsor, primary data analysis, peer-reviewed methodology.

Companion Sources in Corpus

• Federal Reserve FEDS Note (Allen, April 2026): research/07-adoption-challenges/fed-reserve-ai-adoption-monitoring-2026.md — U.S. firm-level adoption benchmarks (18% firm adoption, 41% worker GenAI use).
• DataCamp/YouGov (n=517, February 2026): research/07-adoption-challenges/datacamp-yougov-ai-roi-workforce-capability-2026.md — 2x ROI multiplier from mature training programs; 21% significant ROI baseline.
• PwC Global AI Jobs Barometer 2025: research/07-adoption-challenges/ai-talent-retention-crisis-mid-market.md — 56% wage premium for AI-skilled workers, doubled from 25% in one year.
• ATD 2025 State of the Industry: research/07-adoption-challenges/ai-training-curriculum-by-role.md — $1,254/employee, $165/learning hour benchmarks.
• BCG AI at Work (n=10,635): research/07-adoption-challenges/ai-training-investment-minimum-viable.md — 5-hour training threshold for trust and usage adoption.

See also (wiki): ai-talent-workforce-planning, training-architecture, firm-size-ai-outcomes

Brandon Sneider | brandon@brandonsneider.com April 2026