Which jobs are most exposed to AI?

According to Dario Amodei (Anthropic), mainly entry-level office jobs in finance, law, consulting and tech. The World Economic Forum adds administrative roles, cashiers, bank tellers and postal workers among the occupations in decline by 2030.

Is AI really saving companies money?

Rarely at this stage: MIT estimates that 95% of corporate generative AI projects have produced no measurable return. Where AI does pay, it is mainly the automation of back-office tasks (cost reduction), not the mass replacement of skilled jobs.

How many jobs will AI create or destroy by 2030?

The World Economic Forum forecasts 92 million jobs displaced and 170 million created by 2030, a net gain of 78 million. But that balance hides a mismatch: the jobs destroyed and the jobs created do not involve the same people, the same places or the same skills.

Is "laying off because of AI" sometimes a pretext?

Often, in part. The Harvard Business Review notes that companies are laying off on AI's potential, not on its proven performance. The "AI" reason also serves to justify cost cuts (sometimes inherited from post-Covid over-hiring) and is well received by financial markets.

AI layoffs 2026: the reversal behind the headlines

Q: Are there laws protecting employees against AI?

They are starting to appear: California (executive order N-6-26, May 2026) is revising its layoff rules in the face of AI displacement, and Colorado (AI Act, in force on 30 June 2026) requires employers to prevent algorithmic discrimination in hiring decisions.

Layoffs “because of AI”: the reversal

In 2026, tech is cutting jobs at a record pace — and pointing at a ready-made culprit: artificial intelligence. Except that the figures tell a different story. The productivity gains promised to justify those cuts are, for now, not showing up. A breakdown, without hype.

The figures: a wave of “AI” cuts

The pace is brutal: around 1,115 job cuts a day in tech in 2026, close to double 2025, and some 184,000 jobs cut while citing AI (TechTimes, 16 June 2026). The names are piling up: Meta is laying off 8,000 people (~10%) while redeploying 7,000 employees to AI teams (May 2026); Snap is cutting ~1,000 roles (~16%); Amazon, Oracle and Block are also on the list.

The narrative has become a communications reflex: announce cuts, cite AI, and the market applauds the “efficiency”. Agents able to carry out tasks on their own and tools that write code in place of developers make the argument credible. But credible does not mean verified.

The catch: AI is not (yet) delivering on its promises

That is where the narrative cracks. According to a study by MIT (Project NANDA, “The GenAI Divide”, 2025), 95% of corporate generative AI projects have produced no measurable return on the accounts, despite 30 to 40 billion dollars invested; only 5% genuinely create value. A Gartner study (May 2026, 350 companies) drives the point home: those cutting the most show no improvement in their financial returns (Fortune).

The Harvard Business Review (January 2026) sums up the paradox: companies are laying off on AI’s supposed potential, not on its actual performance. They are betting on a productivity that is not there yet — a bet that AI’s own economics are struggling to fund.

The reversal: they are rehiring

The logical consequence: the backlash. According to Forrester (Future of Work 2026), 55% of employers regret AI-related layoffs, and the firm forecasts that half of those roles will be quietly rehired — often abroad or at a reduced salary. Already, around a third of the companies that tried to replace humans with AI have rehired or reversed course (the “boomerang effect”).

A survey of 2,000 managers explains why: 40% find that AI does not replace in-house know-how, 38% underestimated the need for human oversight, 35% are disappointed by the productivity gains.

What the forecasts say

The projections remain spectacular — and should be handled with care. Dario Amodei, head of Anthropic, warned (May 2025) that AI could wipe out up to 50% of entry-level office jobs and push unemployment to 10-20% within 1 to 5 years (finance, law, consulting, tech). A year later, he is qualifying his own statement (Fortune, May 2026).

At the macro level, the World Economic Forum (Future of Jobs 2025) expects 92 million jobs displaced but 170 million created by 2030, a net +78 million — with one major caveat: these are not the same occupations, the same places or the same skills.

The consequences

Beyond the figures, three effects are taking shape. First a brutal human cost: at Meta, employees in Singapore learned of their dismissal by email at 4 in the morning. Then a destruction of value and knowledge: cutting before AI delivers means losing skills you then have to buy back at a higher price. Finally, precarity: rehiring often happens abroad, at lower pay.

Regulators are starting to react: California (executive order N-6-26, 21 May 2026) is revising its layoff rules in the face of AI displacement, and Colorado (AI Act, in force on 30 June 2026) requires employers to prevent algorithmic discrimination in hiring.

Le Recul’s reading

“Because of AI” has become a magic formula: it justifies cost cuts (sometimes inherited from post-Covid over-hiring) and pleases the financial markets. But the facts sketch a different scenario. In the short term, the layoffs will continue, carried by the narrative. In the medium term, some of them will be quietly reversed. And the real shift, expected around 2030, will play out above all in the pain of retraining.

So the risk is not so much that AI replaces workers. It is that we replace them too early, on a promise not yet kept.

What to take away

~1,115 cuts a day in tech in 2026; 184,000 jobs cut while citing AI (Meta 8,000, Snap ~1,000, Amazon, Oracle, Block).
MIT: 95% of corporate generative AI projects with no measurable return; Gartner: no financial gain for those cutting the most.
Forrester: 55% of employers regret it; half of the “AI” roles would be rehired (often abroad).
Forecasts: Amodei mentions −50% of entry-level jobs / unemployment at 10-20% (1-5 years); WEF: −92 M / +170 M by 2030.

The figure to remember: 95%. That is the share of corporate generative AI projects which, according to MIT, produced no measurable return.

Our verdict. AI promises to replace workers; Le Recul checks: for now, what it mainly replaces is a line in a press release. The gains are not (yet) there, the regrets are piling up, and some of those laid off could soon be called back. To watch: the rehiring wave announced for 2026-2027, and the gap, by 2030, between the jobs destroyed and those actually recreated.

Layoffs "because of AI": the reversal