AI & Tech Intelligence · ILO · OECD · IMF
AI and Jobs: How Many Jobs Are Exposed, and Which Ones
How AI affects jobs, by the numbers: IMF says ~40% of jobs globally are exposed, ILO finds 1 in 4, and most work is transformed, not replaced. Sourced.
TL;DR — The IMF estimates AI could affect about 40% of jobs globally and roughly 60% in advanced economies, either augmenting them or lowering demand for them (IMF, 2024).
ILO–NASK · IMF · OECD Employment Outlook 2023 · WEF | 2,602 words · 16 sections | data: CSV + JSON
Executive summary
Will AI take your job? For most people, the evidence says it will change your job before it eliminates it. The IMF estimates that about 40% of jobs globally — and roughly 60% in advanced economies — are exposed to AI, but exposure cuts both ways: some jobs are made more productive, others face lower demand (IMF, 2024). The ILO's 2025 global index, built with Poland's NASK institute, puts generative-AI exposure at about 1 in 4 jobs worldwide (34% in high-income countries) and reaches a careful conclusion: because most occupations mix automatable and human-only tasks, transformation is far more likely than wholesale replacement (ILO, 2025). The OECD similarly finds 27% of jobs in occupations at highest automation risk on a skills basis, yet reports little evidence so far of AI-driven job losses (OECD, 2023). Exposure is uneven: clerical work is the single most exposed group, and in rich countries women's jobs are disproportionately affected. Looking ahead, the WEF projects a net gain of 78 million jobs by 2030 (170 million created, 92 million displaced) while 39% of core skills shift (WEF, 2025). Field studies show AI mostly augments workers and compresses skill gaps (Brynjolfsson et al., QJE 2025; Stanford HAI, 2026). The honest summary: task-level exposure is broad and real, mass overnight automation is not what the serious literature forecasts, and the outcome depends heavily on reskilling and policy.
“AI will affect almost 40 percent of jobs around the world, replacing some and complementing others.”
Key findings
About 40% of jobs globally are exposed to AI — near 60% in advanced economies
IMF analysis published in January 2024 estimates that roughly 40% of global employment is exposed to AI, rising to about 60% in advanced economies and falling to around 40% in emerging markets and 26% in low-income countries. 'Exposed' is not the same as 'lost': the IMF splits exposed jobs into those where AI raises productivity and those where it lowers labour demand. Managing Director Kristalina Georgieva has warned the shift could hit labour markets 'like a tsunami' and worsen inequality if unmanaged, but the underlying model is explicitly about task exposure, not guaranteed displacement.
Source: IMF · 2024 · confidence: Medium-High
The ILO finds 1 in 4 jobs exposed — and concludes transformation beats replacement
The ILO's May 2025 'Refined Global Index of Occupational Exposure', produced with Poland's NASK, finds about 25% of global employment sits in occupations exposed to generative AI, rising to 34% in high-income countries. Crucially, the ILO stresses that few jobs are fully automatable end-to-end; most are candidates for transformation as AI absorbs specific tasks. Clerical and administrative roles show the highest exposure, and the index measures potential exposure rather than realised job loss, which depends on infrastructure, cost and adoption.
Source: ILO · 2025 · confidence: High
By 2030 the WEF expects a net gain of 78 million jobs, with skills churning fast
The World Economic Forum's Future of Jobs Report 2025, based on over 1,000 employers representing 14 million workers across 55 economies, projects 170 million new roles created and 92 million displaced by 2030 — a net increase of 78 million, equivalent to churning about 22% of jobs. Technology and AI roles are among the fastest-growing, but so are frontline roles like care work and delivery. The report also finds 39% of workers' core skills are expected to change by 2030 (down from 44% in 2023), and that 59% of workers will need reskilling or upskilling. These are employer expectations, not certainties, and carry meaningful forecast uncertainty.
Source: WEF · 2025 · confidence: Medium
2025–2026 evidence tilts toward augmentation, not mass displacement
The most recent primary sources converge on a nuanced picture. The ILO's 2025 index (released May 2025) actually revised exposure toward a more realistic, task-level assessment and emphasised transformation over replacement. The Stanford HAI AI Index 2026 economy chapter highlights a growing body of studies showing AI raises worker output and narrows skill gaps rather than simply removing headcount. And field research such as Brynjolfsson, Li and Raymond's 'Generative AI at Work' (Quarterly Journal of Economics, 2025) documents a 14% average productivity gain for customer-support agents, with the largest gains — up to 34% — accruing to less-experienced workers. The debate is far from settled: the IMF continues to warn about inequality and entry-level exposure, and forecasts of net job creation depend on adoption speed and reskilling that may not materialise evenly.
Will AI take your job? The honest answer
The most defensible answer from the primary research is: AI is more likely to change your job than to eliminate it outright. Exposure is measured at the level of tasks, not whole occupations, and almost every job is a bundle of tasks — some automatable, some not. The ILO's 2025 global index states plainly that 'transformation, not replacement, is the most likely outcome', because occupations mix tasks that generative AI can support with tasks that still require human judgement, accountability and physical presence (ILO, 2025).
That does not mean 'nothing to see here'. The IMF estimates about 40% of jobs globally are exposed, and roughly 60% in advanced economies, precisely because knowledge work — the backbone of rich-country labour markets — is highly susceptible to AI on a task basis (IMF, 2024). The key distinction is between exposure and displacement: a job can be heavily exposed and end up augmented, higher-paid and more productive, or exposed and see demand and wages fall. Which way it breaks depends on adoption, regulation and reskilling.
The OECD reinforces the caution on both sides. It finds 27% of jobs are in occupations at the highest risk of automation on a skills measure, yet reports 'little evidence so far' that AI is causing job losses, and notes that even highly exposed occupations are unlikely to be fully replaced (OECD, 2023). In short: broad task-level exposure is real; mass overnight automation is not what the serious literature forecasts.
How many jobs are exposed? Estimates by source
Headline numbers vary because each institution measures a different thing — 'affected/exposed', 'high automation risk', or 'displaced vs created'. Comparing them side by side is the only honest way to read the field. The table below shows the leading estimates with their scope and year.
The estimates are not contradictory once you read the scope column. The IMF's ~40% is the broadest 'exposure' measure. The ILO's ~25% is a stricter occupational-exposure index. The OECD's 27% is a skills-based 'highest risk' band, not a prediction of losses. The WEF's figures are net labour-market flows to 2030, not exposure at all.
| Source (year) | Estimate | What it measures | Scope |
|---|---|---|---|
| IMF (2024) | ~40% globally; ~60% advanced economies | Jobs exposed to AI (augmented or reduced demand) | Global; by income group |
| ILO–NASK (2025) | ~25% globally; 34% high-income | Occupational exposure to generative AI | Global; potential, not realised |
| OECD (2023) | 27% of jobs | Occupations at highest automation risk (skills-based) | OECD countries |
| WEF (2025) | 92M displaced / 170M created; net +78M | Projected labour-market flows by 2030 | 55 economies, 1,000+ employers |
| IMF (2024) | ~26% low-income economies | Jobs exposed to AI | Low-income countries |
Which occupations are most — and least — exposed
Exposure is highly uneven across occupations. Clerical and administrative work is consistently the single most exposed group in the ILO index, because generative AI is strong at drafting, summarising, scheduling and data entry (ILO, 2025). The OECD adds a striking twist: highly educated white-collar fields such as finance, law and medicine are newly exposed by AI in a way earlier automation waves did not touch, because their core work involves language and pattern recognition (OECD, 2023).
At the other end, jobs dominated by physical dexterity, in-person care, on-site trades and complex human negotiation remain least exposed on current evidence. The table below summarises the pattern reported across the ILO and OECD work. Treat it as directional exposure, not a verdict on any individual job.
Two caveats matter. First, high exposure among skilled professionals usually means task augmentation, not elimination — a lawyer or analyst may do more, faster, rather than disappear. Second, 'lowest exposure' is not 'zero exposure': the OECD notes physical work is no longer fully insulated as AI moves into robotics and logistics.
| Exposure level | Example occupation groups | Primary source |
|---|---|---|
| Highest | Clerical and administrative support; data entry; bookkeeping | ILO (2025) |
| High | Programming, translation/interpretation; parts of finance, law, medicine | OECD (2023) |
| Moderate | Customer service, sales, marketing, analysis (heavy augmentation) | ILO (2025); Brynjolfsson et al. (2025) |
| Lower | Management, community and social services | OECD (2023) |
| Lowest | Skilled trades, construction, in-person care, hospitality, physical/manual work | ILO (2025) |
Augmentation vs displacement: what the evidence shows
The distinction that decides whether AI is a threat or a tool is augmentation versus displacement. The strongest field evidence to date points to augmentation. Brynjolfsson, Li and Raymond's study of 5,179 customer-support agents (Quarterly Journal of Economics, 2025) found access to a generative-AI assistant raised issues resolved per hour by 14% on average — and by up to 34% for novice and low-skilled workers, with minimal effect on the most experienced. AI compressed the skill gap rather than replacing the workforce.
The Stanford HAI AI Index 2026 economy chapter echoes this, citing a growing body of studies where AI increases output and lets less-experienced staff reach expert-level results — 'skill augmentation'. HAI's Erik Brynjolfsson has argued against the 'fallacy' that the only way to get productivity from AI is to cut labour costs; the larger prize is making existing workers more productive (Stanford HAI, 2026).
None of this rules out displacement in specific tasks and entry-level roles. The IMF specifically flags that the tasks most easily eliminated are often those performed in entry-level jobs, which has implications for how young workers get on the ladder (IMF, 2024). The realistic model is: many jobs augmented, some tasks automated away, and pressure concentrated at the entry level and in the most exposed clerical roles.
The jobs AI is creating
Every major forecast pairs displacement with creation. The WEF's Future of Jobs Report 2025 projects 170 million new roles by 2030 against 92 million displaced — a net gain of 78 million (WEF, 2025). The fastest-growing roles include AI and machine-learning specialists, data analysts and scientists, and other technology positions, but the report stresses that growth is not confined to tech: frontline and 'core economy' roles such as care workers, delivery drivers, educators and farmworkers also expand.
The demand signal is shifting toward AI-complementary skills. The WEF reports that analytical thinking remains the most sought-after skill (cited by 70% of employers), and that AI and big-data skills are among the fastest-rising. In high-AI-exposure occupations, demand is rising for uniquely human capabilities — the OECD documents growing demand for originality and social skills in exactly the jobs most exposed to AI, with Sweden among the countries showing the largest such increases.
The catch is timing and matching. New roles do not automatically go to displaced workers, which is why the WEF finds 59% of the workforce will need reskilling or upskilling by 2030 — and warns that about 11 in every 100 workers are unlikely to receive the training they need (WEF, 2025).
Productivity and wages: the double-edged effect
AI's labour-market impact runs through productivity, and the evidence is genuinely positive on output. Beyond the 14% support-agent gain, the Stanford HAI AI Index 2026 documents measurable productivity improvements across writing, coding and analytical tasks, typically largest for less-experienced workers (Stanford HAI, 2026). Higher productivity can translate into higher pay for augmented workers — the IMF notes that in advanced economies some workers whose jobs are enhanced by AI have already seen higher incomes (IMF, 2024).
But the same mechanism can widen inequality. The IMF's central warning is that AI may worsen overall inequality: high-income, high-skill workers who can harness AI stand to gain, while workers whose tasks are substituted may face lower demand and wages. Where the gains land — labour or capital, senior or junior, augmented or displaced — is a distributional question that policy, not technology alone, will decide.
This is why serious analysts avoid a single headline number for 'the wage effect'. The honest reading is a fork: augmentation raising productivity and pay for many, alongside downward pressure and displacement risk for the most substitutable tasks, with the net distributional outcome still uncertain.
Who is most affected: clerical workers and women in rich countries
Exposure is not evenly distributed across people. The clearest demographic finding comes from the ILO's 2025 index: women's employment is disproportionately exposed in high-income countries, because women are over-represented in clerical and administrative roles — the single most exposed group. In high-income countries, 9.6% of female employment falls in the highest-exposure category versus 3.5% of male employment; globally the split is 4.7% of women versus 2.4% of men (ILO, 2025).
The IMF adds an age dimension: because entry-level jobs concentrate the tasks most easily automated, younger workers at the start of their careers may be more exposed to task displacement, even as experienced workers are more likely to be augmented (IMF, 2024). The OECD's earlier skills-based analysis found the highest shares of automatable skills among lower-skilled and younger workers, though its AI-specific findings extend exposure up the skills ladder into professional occupations (OECD, 2023).
The policy implication is targeting. A blanket 'reskill everyone' message misses that exposure is concentrated — by task, by gender, by income level and by career stage — and that mitigation should follow the same contours.
The Nordic and European angle
Advanced European economies, including the Nordics, sit near the top of the exposure range precisely because their labour markets are knowledge- and services-heavy. The ILO's 34% exposure figure for high-income countries applies directly to Sweden, Denmark, Norway and Finland, whose large shares of clerical, administrative and professional work are the most AI-exposed categories (ILO, 2025).
The OECD's work on skills in the AI age finds that Sweden is among the countries with the largest increases in demand for originality and social skills within high-AI-exposure occupations — a sign that European employers are already steering exposed roles toward human-complementary tasks rather than cutting them (OECD, 2024–2026). EU-level analysis from the Commission's Joint Research Centre similarly frames AI as reshaping task content across the single market, with clerical roles most affected but often retrainable.
The Nordic advantage is institutional: strong adult-education systems, active labour-market policy and high trust make task transformation easier to absorb than in economies with weaker reskilling infrastructure. That is a mitigating factor, not immunity — the exposure share is high, so the burden falls on whether reskilling keeps pace.
What workers and policymakers should do
For workers, the evidence points to a clear strategy: learn to use AI as a tool rather than compete with it. The productivity studies show the largest gains accrue to people who adopt AI to augment their work, and the WEF finds analytical thinking, AI literacy, resilience and flexibility among the most valued skills to 2030 (WEF, 2025; Stanford HAI, 2026). Moving up the task ladder — toward judgement, relationships, originality and accountability that AI cannot own — is the durable hedge.
For policymakers, the shared message across the IMF, ILO and OECD is to manage the transition actively so productivity gains are broadly shared. That means large-scale reskilling and upskilling (the WEF estimates 59% of workers need it), social-protection systems that adapt to changing work, and targeted support for the most exposed groups — clerical workers, women in high-income countries and entry-level staff (IMF, 2024; ILO, 2025).
The framing that unites the serious literature: AI's labour outcome is not predetermined by the technology. It is a policy and adoption choice about whether exposure becomes augmentation or displacement.
Limitations and uncertainty of the forecasts
These numbers deserve honest caveats — treating them as precise predictions would be a mistake. 'Exposure' indices (IMF, ILO, OECD) measure potential based on task content; they do not predict how many jobs will actually change, which depends on cost, infrastructure, regulation and adoption speed that vary widely by country and sector (ILO, 2025).
Forecasts of net job flows carry even more uncertainty. The WEF's 170-million/92-million figures are aggregated employer expectations from a survey, not observed outcomes, and the report itself notes projections shift year to year (the share of skills expected to change fell from 44% in 2023 to 39% in 2025). Historically, technology forecasts have often mis-timed and mis-sized both the losses and the gains.
The bottom line for a careful reader: the direction of the evidence is robust — broad task-level exposure, augmentation as the dominant near-term effect, concentrated risk for clerical and entry-level work — but the precise magnitudes and timing remain genuinely uncertain. That uncertainty is itself a finding, and any source claiming precision about how many jobs AI will 'take' is overstating what the data supports.
Scoreboard (machine-readable data)
Every headline indicator with its value, period, source and confidence. Free to reuse under CC BY 4.0.
| Indicator | Value | Period | Source | Conf. |
|---|---|---|---|---|
| Global jobs affected by AI | 40 pct | 2024 | IMF | Medium-High |
| Advanced-economy jobs affected | 60 pct | 2024 | IMF | Medium-High |
| Jobs exposed to generative AI | 25 pct | 2025 | ILO | High |
| High-income jobs exposed to generative AI | 34 pct | 2025 | ILO | High |
| Jobs at highest automation risk | 27 pct | 2023 | OECD | Medium-High |
| Net new jobs by 2030 | 78 millions | 2025 | WEF | Medium |
Methodology & verification
This report synthesises the leading primary-source estimates of AI's labour-market impact and reports each figure with its named source, year and the specific concept it measures. We deliberately do not blend incompatible measures into a single number: 'exposure' (IMF, ILO), 'highest automation risk' on a skills basis (OECD), and projected net labour-market flows (WEF) are reported separately with scope noted. Exposure indices describe potential based on task content, not realised job losses. Field-study evidence on productivity (Brynjolfsson, Li & Raymond, QJE 2025) and the Stanford HAI AI Index 2026 are used to characterise augmentation versus displacement. Every statistic is attributed to a primary or peer-reviewed source; where sources disagree or measure different things, we present the range rather than a point estimate. Figures were checked against source publications as of 30 July 2026.
Data dictionary
| Field | Type | Description |
|---|---|---|
| jobs_exposed_pct | percentage | Share of employment in occupations exposed to AI/generative AI (potential task exposure, not realised job loss), per the cited source and geography. |
| jobs_high_risk_pct | percentage | OECD skills-based measure: share of jobs in occupations at highest risk of automation (occupations with more than 25 of 100 skills experts deem easily automatable). |
| net_jobs_2030_millions | number (millions) | WEF projected net change in jobs by 2030 (roles created minus roles displaced), aggregated from employer survey expectations. |
Frequently asked questions
Which jobs are most at risk from AI?
Clerical and administrative roles are the single most exposed group, according to the ILO's 2025 global index, because generative AI is strong at drafting, data entry, scheduling and summarising. The OECD adds that parts of programming, translation, finance, law and medicine are also newly exposed. 'At risk' mostly means task transformation, not automatic elimination (ILO, 2025; OECD, 2023).
Will AI take my job?
For most people, AI is more likely to change your job than eliminate it. The ILO concludes transformation, not replacement, is the most likely outcome because jobs bundle automatable and human-only tasks. Exposure is broad — the IMF puts it at ~40% of jobs globally — but exposure includes jobs that AI makes more productive, not just jobs it removes (ILO, 2025; IMF, 2024).
How many jobs will AI create versus destroy?
The WEF's Future of Jobs Report 2025 projects 170 million new jobs created and 92 million displaced by 2030 — a net gain of 78 million across 55 economies. These are employer expectations, not certainties, and the fastest-growing roles include both AI/tech specialists and frontline care and delivery work (WEF, 2025).
Is AI more likely to help or replace workers?
Current field evidence leans toward help (augmentation). A study of 5,000+ customer-support agents found a 14% average productivity gain, rising to 34% for less-experienced staff, with AI compressing skill gaps rather than replacing workers. The Stanford HAI AI Index 2026 reports similar augmentation across writing, coding and analysis (Brynjolfsson et al., QJE 2025; Stanford HAI, 2026).
Are women more exposed to AI at work?
In high-income countries, yes. The ILO finds 9.6% of female employment falls in the highest-exposure category versus 3.5% of male employment, because women are over-represented in clerical and administrative roles. Globally the gap is narrower (4.7% of women versus 2.4% of men) (ILO, 2025).
How reliable are these AI job forecasts?
Directionally reliable, numerically uncertain. Exposure indices measure potential based on task content, not actual job losses, and net-flow forecasts are aggregated survey expectations that shift year to year. The robust findings are broad task exposure, augmentation as the dominant near-term effect, and concentrated risk for clerical and entry-level work; precise magnitudes and timing are not settled (ILO, 2025; WEF, 2025).
Glossary
- Task exposure
- The degree to which the individual tasks that make up a job can be performed or supported by AI. Most exposure measures work at the task level, since occupations rarely automate as a whole. ↗
- Augmentation
- When AI increases a worker's productivity or capability rather than replacing them — for example, letting a novice reach expert-level output. The dominant near-term effect in field studies. ↗
- Displacement
- When AI substitutes for the labour previously required to perform a task or role, reducing demand for that work. Concentrated in easily automatable tasks and often entry-level jobs. ↗
- Highest automation risk
- The OECD's skills-based band for occupations with more than 25 of 100 skills experts consider easily automatable; about 27% of jobs. It denotes risk exposure, not confirmed job loss. ↗
Embed & cite this report
Free to reuse under CC BY 4.0. Embed the live-updating widget on your site, or cite the report directly — always with attribution to Affärslivet.
Embed (HTML) — auto-updating
<iframe src="https://xn--affrslivet-s5a.com/en/embed/ai-jobs-impact" width="100%" height="520" style="border:1px solid #e3e3e6" title="AI and Jobs: How Many Jobs Are Exposed, and Which Ones — Affärslivet" loading="lazy"></iframe> <p style="font:12px sans-serif">Source: <a href="https://xn--affrslivet-s5a.com/en/reports/ai-jobs-impact">Affärslivet</a></p>
APA
Affärslivet Research. (2026). AI and Jobs: How Many Jobs Are Exposed, and Which Ones. Affärslivet. Version 1.0. https://xn--affrslivet-s5a.com/en/reports/ai-jobs-impact
MLA
Affärslivet Research. "AI and Jobs: How Many Jobs Are Exposed, and Which Ones." Affärslivet, 2026-07-30, https://xn--affrslivet-s5a.com/en/reports/ai-jobs-impact.
BibTeX
@techreport{affarslivet_ai_jobs_impact,
title = {AI and Jobs: How Many Jobs Are Exposed, and Which Ones},
author = {{Affärslivet Research}},
year = {2026},
note = {Version 1.0},
url = {https://xn--affrslivet-s5a.com/en/reports/ai-jobs-impact}
} License CC BY 4.0 — free to cite, embed and republish with attribution to Affärslivet. Data also as CSV / JSON.
Sources
- ILO–NASK — Generative AI and Jobs: A Refined Global Index of Occupational Exposure (2025)
- IMF — AI Will Transform the Global Economy / Gen-AI: Artificial Intelligence and the Future of Work (2024)
- OECD Employment Outlook 2023 — Artificial Intelligence and Jobs
- WEF — Future of Jobs Report 2025
- Stanford HAI — 2026 AI Index Report, Economy chapter
- Brynjolfsson, Li & Raymond — Generative AI at Work, Quarterly Journal of Economics (2025)
Part of Affärslivet AI Intelligence
This report is one part of Affärslivet's source-cited AI knowledge layer. Start with the big picture:
- The State of AI 2026 — our free, source-cited annual report
- AI Statistics 2026 — 230+ source-cited AI facts & figures
- The AI Intelligence hub — all our indexes, rankings, model & lab profiles