IMF: AI could affect 40% of jobs. What the number means, and why exposure isn't job loss

On Monday, October 12, 2026, the opening day of the IMF and World Bank Annual Meetings in Bangkok, IMF Managing Director Kristalina Georgieva told a panel on AI and employment that AI is "a slow-moving tsunami, but [a] tsunami nonetheless." Reports from the meetings repeated the IMF's best-known statistic: AI could affect about 40% of jobs worldwide, rising to about 60% in advanced economies. The figure is real and comes from IMF research. It is also nearly three years old, and it measures exposure, which is not the same as how many jobs will disappear. This analysis traces the number to its source, explains what it measures, and sets out what the IMF's newer evidence adds. The estimates are about the capabilities of frontier models reaching the workplace.

This analysis is based on the IMF's January 2024 Staff Discussion Note and its January 2026 follow-up, IMF blog posts by Georgieva, IMF Research Department slides, and Reuters and other reporting from Bangkok, all checked October 12, 2026.

The short version

• Source of the 40%: IMF Staff Discussion Note 2024/001, Gen-AI: Artificial Intelligence and the Future of Work, published January 14, 2024.
• The numbers: almost 40% of global employment is exposed to AI. By income group: about 60% in advanced economies, 40% in emerging markets and 26% in low-income countries.
• What "exposed" means: a job's tasks overlap with what AI can do. It is not a forecast of layoffs.
• The split that matters: in advanced economies, roughly half of exposed jobs are likely to be complemented by AI and half face higher displacement risk.
• What's new: the IMF's January 2026 note finds early measurable effects on skills, wages and employment in vulnerable occupations.

What did Georgieva say in Bangkok?

Reuters reported from Bangkok that Georgieva described AI as "a transformative technology" that "so completely changes how we work, how we live," and said: "AI, love it, hate it. It's already here." The Irish Times reported the "slow-moving tsunami" line. Georgieva said, "We cannot today imagine what exactly the world of work will be five years from now." She also warned that AI could strain tax revenues by shifting income from labor to capital.

According to Reuters, the IMF also says strong AI investment is helping keep global growth near 3%, and it estimates AI could add 0.5 to 1 percentage point to growth in advanced Asian economies. The investment side of that story is the datacenter build-out we track in the gigawatt datacenter race. An updated IMF growth outlook was due on Tuesday, October 13.

The jobs figures in the coverage are reported speech or IMF estimates, not verbatim quotes. Reuters wrote that IMF research suggests about 40% of jobs will be "augmented, downgraded, or eliminated by AI," rising to as much as 60% in advanced economies. One wire story (from Nigeria's News Agency, republished by several outlets) gives the low-income figure as 23%. The IMF's published estimate is 26%, and we use that.

Where does the 40% come from?

The number was first published on January 14, 2024, in IMF Staff Discussion Note 2024/001 by Mauro Cazzaniga, Florence Jaumotte, Longji Li, Giovanni Melina, Carlo Pizzinelli, Emma J. Rockall and Marina Mendes Tavares. Georgieva announced it the same day in an IMF blog post:

"almost 40 percent of global employment is exposed to AI."

She added that in advanced economies "about 60 percent of jobs may be impacted by AI," and that in emerging markets and low-income countries "AI exposure is expected to be 40 percent and 26 percent, respectively."

Two years later, her January 14, 2026 blog repeated the headline: "Nearly 40 percent of global jobs" are exposed to AI-driven change. That's why the figure keeps returning. It has stayed the IMF's reference number even as the underlying models have advanced.

What does "exposed to AI" actually mean?

This is the part most headlines skip. According to the IMF Research Department's presentation of the study, exposure is measured with the AI Occupational Exposure (AIOE) index developed by Felten and coauthors. It captures the "degree of overlap between AI applications and human abilities in occupations." A job counts as highly exposed if its occupation falls in the top 30% of AIOE scores. Shares are weighted by hours worked.

Exposure says that AI could plausibly do or assist with much of the work in a job. It says nothing about:

• whether employers will actually adopt AI for those tasks, or when;
• whether adopting it is cheaper than the people doing the work today;
• whether regulation, liability or customers will allow it;
• whether demand for the work rises when it gets cheaper.

So exposure is an upper-bound map of where change could happen, not a count of jobs that will go. Reuters' paraphrase of jobs being "augmented, downgraded, or eliminated" covers all three outcomes, and the IMF's own work treats them separately.

Complement or replace? The split inside the 60%

The IMF's distinctive step was to add a complementarity index (the researchers also call it a "shielding" index). It is built from O*NET data on each job's work context and skills: how much communication and responsibility it involves, physical conditions, how critical its decisions are, how routine it is, and the skill level required. A highly exposed job with high complementarity is more likely to be augmented. The IMF's example is judges. A highly exposed job with low complementarity faces higher displacement risk. The example is clerical workers.

Here is the breakdown, from the IMF Research Department's slides presenting SDN 2024/001:

Last verified: October 12, 2026 (IMF, SDN 2024/001 and the Research Department's May 2024 slides)

Georgieva's 2024 blog summed it up: "Roughly half the exposed jobs may benefit from AI integration, enhancing productivity." For the other half, "AI applications may execute key tasks currently performed by humans, which could lower labor demand, leading to lower wages and reduced hiring. In the most extreme cases, some of these jobs may disappear."

Even the "higher risk" third of advanced-economy jobs is a risk category, not a forecast. The IMF doesn't say what share of that 33% will actually be cut.

Why are poorer countries less exposed, and is that good news?

Not exactly. Low-income countries have more employment in agriculture and manual work, which language and vision models don't reach yet, so fewer of their jobs are exposed. The IMF's 2024 analysis warned that these countries also "don't have the infrastructure or skilled workforces to harness the benefits of AI," which could widen gaps between countries over time. Coverage from Bangkok made the same point: the lower figure reflects lower AI adoption, not lower risk to individual workers. The IMF also concluded that "in most scenarios, AI will likely worsen overall inequality," and that "younger workers may find it easier to exploit opportunities, while older workers could struggle to adapt."

What has changed since 2024?

The IMF's newer evidence moves from potential exposure to observed effects. Staff Discussion Note 2026/001, Bridging Skill Gaps for the Future: New Jobs Creation in the AI Age, was presented in Georgieva's January 14, 2026 blog. It reports:

• New skills are spreading. About "one in 10 job postings in advanced economies" requires at least one new skill, compared with one in 20 in emerging markets.
• New skills pay more. Postings with at least one new skill pay about 3% more in the UK and US. Postings with four or more new skills pay up to 15% more in the UK and 8.5% more in the US.
• Some displacement shows up in the data. In regions with high demand for AI skills, employment in AI-vulnerable occupations is "3.6 percent lower after five years" than in regions with less demand.
• Entry-level hiring is under pressure. The blog cites emerging US evidence that "generative AI adoption reduces entry-level hiring," especially for automatable tasks.

The IMF's 2026 Annual Report adds that middle-skilled workers in highly exposed jobs "are being left out," while cities and regions with more AI-related jobs "are not experiencing overall job growth." Together these findings suggest real but gradual effects, which fits the "slow-moving tsunami" framing: concentrated in specific occupations and among young entrants, not a sudden collapse in total employment.

There's a further caution on productivity. Exposure estimates assume AI can do the work well. Evidence on that is mixed even in software, where AI tools are most mature. Our explainer on SWE-bench, Terminal-Bench and METR covers why measured productivity gains have been hard to pin down. Much of today's change also runs through autonomous tools rather than chatbots, which our explainer on what "agentic" really means describes.

What does the IMF want governments to do?

The 2024 note called on countries to "establish comprehensive social safety nets and offer retraining programs for vulnerable workers." Its AI Preparedness Index ranked countries on digital infrastructure, human capital and labor-market policy, innovation, and regulation. The 2026 note adds a Skill Imbalance Index and splits the advice by economy. Where demand for new skills outpaces supply, it recommends education and reskilling. Where the supply of skills outpaces demand, it recommends helping firms absorb workers through innovation and access to credit.

What's confirmed, reported and uncertain

• Confirmed (IMF documents): the 40% / 60% / 40% / 26% exposure estimates and their January 14, 2024 source; the exposure and complementarity method and the 27/33 split for advanced economies; and the 2026 findings on new skills, wages and the 3.6% employment effect.
• Reported (Reuters, The Irish Times, NAN/Citi Newsroom): Georgieva's October 12, 2026 remarks in Bangkok, the "slow-moving tsunami" quote, and the IMF's Asian growth estimates.
• Uncertain: how many exposed jobs will actually be cut, augmented or created. The IMF hasn't published a job-loss forecast, and its exposure estimates haven't been updated for 2026-era models.