AI IS WIPING OUT MILLIONS OF JOBS — AND EXPERTS SAY IT'S ABOUT TO GET MUCH WORSE

WASHINGTON — Artificial intelligence is coming for your job.

That, at least, is the warning from a landmark report published this week, which found that AI systems could replace the equivalent of 300 million full-time workers globally — a displacement so vast it would dwarf every previous technological disruption in recorded history.

"We are standing at a precipice," said the report's lead author, Dr. Elena Marchetti, of the Institute for Global Economic Policy. "The tools being deployed right now are categorically different from anything we have seen before. The question is not whether jobs will disappear — it is how many, and how fast."

The findings sent shockwaves through financial markets on Tuesday, with stocks in recruiting, staffing, and administrative-services sectors falling sharply. Labour unions across twelve countries called for emergency sessions.

The Numbers

The report identifies three broad categories of work at "extreme risk": administrative and clerical roles, certain legal and financial processing tasks, and customer-facing service jobs where transactions are routine.

"Those 300 million jobs represent roughly 18 percent of the global formal workforce," said Marchetti. "Many of them are concentrated in wealthy countries, precisely because wealthy countries have invested more heavily in the infrastructure that makes automation possible."

Other analyses reached similarly alarming conclusions. A McKinsey Global Institute study estimated that between 400 and 800 million individuals could be displaced by 2030. The World Economic Forum projects that AI and automation will eliminate 85 million jobs across 26 countries before the end of this decade — a figure that has already begun to materialise in data entry, basic coding, and call-centre work.

The numbers, taken at face value, are staggering. But they come with an important qualifier — one that tends to get lost in the headlines.

The Part Nobody Quotes

"Every single one of these reports also projects significant job creation," says Dr. James Osei, a labour economist at the London School of Economics who reviewed several major studies for this article. "The Goldman Sachs report that everybody quotes? It also says AI could raise global GDP by 7 percent and lift real wages for many workers. The WEF report? It estimates 97 million new jobs created alongside the 85 million displaced. People just don't tend to share that part."

This pattern — alarming displacement numbers paired with less-noticed creation projections — runs through nearly all serious economic analyses of AI's labour market impact. The question is not simply how many jobs disappear, but how many new ones appear, how long the transition takes, and — critically — whether they materialise in the same places and for the same people.

"Transitions in the labour market are real and they hurt," says Osei. "If you are a 55-year-old data entry clerk, 'there will be new jobs eventually' is cold comfort. But if you are asking whether AI will lead to permanent mass unemployment across the whole economy — the historical record suggests that is probably not what happens."

Not All Jobs Are Equal

The risk of displacement is also far from evenly distributed. A closer look at the data reveals that broad aggregate figures can obscure enormous variation in who, specifically, is affected.

A 2024 Brookings Institution analysis found that automation risk correlates strongly with income — but in a counterintuitive direction. It is mid-wage, routine-heavy jobs — not the lowest-wage ones — that are most vulnerable. Manual work requiring physical dexterity and contextual judgment (cleaning, care work, construction) is harder to automate than cognitive work that follows predictable patterns (data processing, basic legal research, scheduling).

Geography matters too. Workers in areas with fewer alternative employment options face greater disruption than those in urban labour markets with more diverse opportunities. The effect in the Global South may unfold very differently than in OECD countries, where social safety nets and retraining programmes can buffer the transition.

"When we say '300 million jobs at risk,' we are aggregating across enormously different situations," says Priya Nair, a researcher at Oxford's Future of Work programme. "A customer service agent in Manila and an accounts payable clerk in Milwaukee are both in that number — but the economic context, the safety net, the transition support available to them: none of that is the same."

We Have Been Here Before

It is worth stepping back, because the current anxiety about AI and employment is not new. It is, in fact, a very old fear.

When mechanical looms were introduced in the early 19th century, the Luddite movement — whose name has since become shorthand for technophobia — destroyed machinery out of genuine economic desperation. Those workers were not wrong that their specific jobs were disappearing. But the broader economy added far more employment in new industries than it destroyed.

The introduction of the ATM in the 1970s was widely predicted to eliminate bank tellers. Instead, teller employment rose for decades afterwards — because ATMs reduced the cost of operating a branch, which meant banks opened more of them, and tellers shifted toward more complex customer-facing tasks that machines could not replicate.

None of this guarantees AI will follow the same trajectory. It may well be genuinely different this time. But it is worth noting that "this time is different" has been said about every major automation wave in history — and has been wrong about the macro picture each time, even as it was right about the specific pain experienced by particular groups of workers in particular places.

What We Actually Know

So what does the evidence actually support?

We know that some jobs are already being eliminated. We know that AI is capable of performing a wider range of tasks than previous technologies — including some that were long assumed to require human judgment. We know the pace of development is fast, and accelerating.

We also know that economic forecasting is hard. The economists who predicted ATMs would eliminate bank tellers were not foolish — they were reasoning carefully from the evidence available at the time. The economists projecting catastrophic AI displacement today may be right. They may be wrong. Most of them will tell you, if pressed, that the genuine uncertainty is large.

"I think there are two things we can say with confidence," says Osei. "First: some people are going to be seriously hurt by this transition, and policy responses that help them adapt are not optional — they are necessary. Second: anyone who claims certainty, in either direction, is probably overestimating their ability to predict the future."

The headline, in other words, is always simpler than the story.