92 Million Jobs at Risk, 170 Million New Ones: The Real Balance Sheet of AI

92 million jobs displaced. 170 million created. Net balance: +78 million. According to the World Economic Forum, AI will generate more work than it destroys by 2030.

That’s what the headlines say. And technically, it’s correct.

But it hides three problems no one wants to address: the jobs that disappear are not the same ones being created, reskilling takes years while automation takes months, and those who suffer most are those with the least capacity to adapt.

The four numbers everyone gets wrong

There are four major studies on AI and employment. All four are constantly cited. And all four are misinterpreted.

WEF: 92 million displaced, 170 million created

The Future of Jobs Report 2025 from the World Economic Forum, published in January 2025, surveyed more than 1,000 employers representing 14 million workers across 55 economies. Its projection for 2030: 22% of current jobs will undergo structural transformation. Of those, 92 million will be displaced and 170 million will be created.

The fastest-growing roles: big data specialists (+113%), fintech engineers (+93%), AI/ML specialists (+82%), software developers (+57%).

The fastest-declining: postal clerks (-34%), bank tellers (-31%), data entry operators (-26%), retail cashiers (-20%), administrative assistants (-20%).

New in 2025: for the first time, graphic designers appear among the fastest-declining roles, attributed directly to generative AI.

IMF: 40% of global jobs exposed

The International Monetary Fund report (January 2024) found that nearly 40% of global employment is exposed to AI. In advanced economies, it rises to 60%. In emerging economies, it drops to 40%. In low-income countries, to 26%.

The key word is exposed — not eliminated. Roughly half of exposed jobs could benefit from AI (higher productivity); the other half could see their core tasks performed by machines.

Kristalina Georgieva, IMF Managing Director, was more blunt at Davos 2026: she described the impact as a “tsunami”, warning that entry-level jobs are disappearing and the middle class is being “squeezed.”

Goldman Sachs: 300 million jobs affected, not eliminated

The Goldman Sachs report (March 2023) is probably the most misinterpreted. Its figure of 300 million jobs refers to jobs affected by automation, not eliminated. Most are “only partially exposed” and are “more likely to be complemented than substituted.” Of the 18% of global work that could be automated, most will be transformed, not disappear.

AI could drive a 7% increase in global GDP (~$7 trillion) over a decade and boost productivity by 1.5 percentage points annually for 10 years.

McKinsey: 57% of work hours are automatable (technical potential)

McKinsey Global Institute (November 2025) found that AI agents could perform tasks occupying 44% of current U.S. work hours. Adding robots, the figure reaches 57%.

Crucial nuance: McKinsey explicitly clarifies that this is “technical potential, not a forecast of job loss.” The fact that a task can be automated doesn’t mean it will be automated, nor that the employee will disappear.

All the big numbers on AI and employment measure exposure or technical potential, not guaranteed destruction. But industry and media cite them as verdicts.

The layoffs already happening

While the reports speak in the future tense, layoffs are already happening in the present.

In 2025, more than 55,000 layoffs in the U.S. were explicitly attributed to AI. In the first six weeks of 2026, the figure reached 61,000. An NBER/Duke-Federal Reserve survey of chief financial officers projects ~502,000 AI-related layoffs for all of 2026 — 9x more than in 2025. 44% of surveyed CFOs plan cuts tied to automation.

More than 45 CEOs have publicly cited AI as a driver of their workforce reduction decisions.

But there’s an important nuance: HBR published an analysis in January 2026 arguing that companies are laying off “because of AI’s potential, not its performance.” Many of these cuts would have happened regardless — AI serves as a convenient justification for restructurings driven by other motivations.

502,000 layoffs sounds alarming. But they represent 0.4% of total U.S. employment (~125 million jobs). As John Graham, co-author of the study, noted: “It’s not the apocalyptic scenario you sometimes see in headlines.”

The silent disaster: freelancers

Where the impact is already measurable and devastating is in the freelance market.

A study published in the Journal of Economic Behavior & Organization (2024), analyzed by the Brookings Institution, found that freelancers in AI-exposed occupations experienced a 2% drop in monthly contracts and a 5% drop in total monthly income in the 6-8 months following the launch of AI tools.

Counterintuitively, more skilled freelancers suffered more than less skilled ones.

The sector-level data is devastating:

• Writing/content: freelance job postings fell 33%; writing projects on Upwork dropped 32% year-over-year in 2025 — the largest decline of any category
• Translation: job postings fell 19%
• Graphic design: job postings dropped 18.5%; 3D modeling 15.6%
• Software development: dropped 21%
• Customer support: gig postings fell 16%

More than half of the companies spending on freelance platforms in 2022 had stopped doing so entirely by 2025. Spending on freelancers as a percentage of total spending fell from 0.66% to 0.14%. Simultaneously, spending on AI models rose from zero to 2.85%.

Rates have bifurcated: freelancers working with AI earn 44% more per hour; those competing against it have seen their rates drop 30-40%.

What the net balance doesn’t tell you

The WEF’s +78 million hides three mismatches:

1. Skills mismatch

A laid-off bank teller doesn’t become a big data specialist. A data entry operator doesn’t become a fintech engineer. The 92 million jobs disappearing require basic or intermediate training. The 170 million being created require STEM, digital, or specialized skills that take years to acquire.

2. Geographic mismatch

The jobs created are concentrated where there is digital infrastructure, universities, and tech ecosystems. The jobs destroyed are spread across the entire economy, including rural areas and developing countries. The IMF quantifies it: the most affected countries are Hong Kong, Israel, Japan, Sweden, and the U.S.; the least affected, mainland China, Nigeria, Vietnam, Kenya, and India.

3. Temporal mismatch

Automation advances at the speed of software. Professional reskilling advances at the speed of the education system. Between the two, there is a 3-7 year gap that no transition plan is covering at the necessary scale.

AI is not a substitute for human beings. It is a productivity tool that can empower them. But that only happens when it is deployed with a transition plan that no one is funding at the scale needed.

What can I do?

• If you’re worried about your job: Evaluate which part of your work is automatable and which part requires judgment, human connection, or original creativity. Invest in the latter. AI fluency is the fastest-growing skill in U.S. job postings — it has multiplied 7x in two years according to McKinsey.

• If you run a company: Before automating, measure the real impact. 95% of organizations are not getting ROI from generative AI. Laying off workers because of “AI’s potential” when you haven’t even proven its value means replacing people with a promise. And European law (CSRD, EU AI Act) will require you to justify the social impact of your technology decisions.

• If you’re a freelancer: Specialize. Freelancers working with AI as a tool earn 44% more. Those competing against it on commoditized tasks lose 30-40%. The difference is human judgment that AI doesn’t replace.

• If you work in policy: Reskilling plans exist on paper. Funding at scale does not exist in practice. 92 million people need transition before 2030. The clock is already ticking.