Frequently asked questions
Will artificial intelligence really eliminate millions of jobs ?
Which jobs are most threatened by AI ?
Can AI also create jobs ?
Should you train in AI to protect your job ?
In late February 2026, a report from the American firm CitriniResearch briefly rattled financial markets. Its scenario? Within two years, artificial intelligence would have rendered such a massive portion of white-collar workers jobless that the global economy would collapse under its own weight.
The text was written in the first person from a fictional future, "June 30, 2028, the unemployment rate stood at 10.2%...", and its dystopian tone was quickly deemed "unconvincing," even "problematic" by many economists.
But it had the merit of raising the question aloud, a question many have been silently asking since the explosion of generative AI.
Two centuries ago, English workers smashed weaving looms with sledgehammers, believing they would condemn them to destitution; this Luddite movement has gone down in history as the symbol of all technological resistance.
Yet, the Industrial Revolution did not produce the mass unemployment that was predicted, but it transformed jobs, eliminating some and creating others, often in much greater numbers.
Economic history repeats this pattern with disturbing regularity: agriculture employed 42% of the American workforce in 1900; a century later, that figure had dropped to 2%.
Yet, the unemployment rate in the United States did not skyrocket as a result; jobs disappeared, others emerged, 60% of which simply didn't exist eighty years ago.
Ford's assembly line, which reduced car assembly time from twelve hours to an hour and a half, also sparked catastrophic predictions; the reduction in car prices actually stimulated demand and employment.
Long-term unemployment data from G7 countries confirms what economic historians have observed for decades: unemployment follows economic cycles much more closely than it follows technological innovation.
Saying "it's always worked out before" isn't enough to close the debate, because generative AI presents a fundamental difference compared to previous revolutions: weaving looms replaced physical strength, computers automated repetitive and codifiable tasks, and AI, for its part, now tackles cognitive tasks—writing, financial analysis, accounting, customer service, computer code—areas previously considered sacrosanct.
OpenAI estimated in 2023 that 80% of the American workforce held jobs where at least 10% of tasks could be performed by AI, and that for 20% of jobs, half of the tasks would be affected.
The World Economic Forum, for its part, predicted that by 2027, AI would create 69 million new jobs globally, while eliminating 83 million, resulting in a net contraction of 14 million positions—figures that give pause for thought, even if they remain projections.
But we must also measure the gap between theoretical potential and actual adoption; according to a PwC report published in May 2025, only 4% of companies believed they were currently deriving significant benefits from AI, and while technology is advancing rapidly, its integration into organizations is slow, costly, and often partial.
The nuance that most serious economists advocate is not "AI will change nothing" but rather "AI will transform jobs more than it will eliminate them."
The OECD emphasizes this distinction: in the majority of observed cases, the adoption of AI in businesses leads to a redeployment of employees to other tasks, not outright dismissal.
Indeed's data for France in April 2026 clearly illustrates this trend: in a generally declining job market, job postings mentioning AI are growing against the trend, and already account for 21% of listings in IT development, 15% in systems and network administration, and 12% in banking and finance.
AI doesn't eliminate these jobs; it reshapes their contours.
The least exposed professions remain those that mobilize creativity, empathy, and judgment in complex and non-standardized situations: a social worker, a surgeon in an operating room, a negotiator, an artist—these profiles combine skills that AI can assist but not autonomously replicate.
Conversely, highly routine functions in accounting, logistics, document processing, or tier-1 customer service are exposed to increasing and rapid automation.
There's a blind spot in this debate, rarely mentioned: AI sometimes serves as a convenient pretext.
As France Info noted in March 2026, AI has become "a formidable pretext for justifying mass layoffs when companies want to cut costs."
Restructurings that would have happened anyway find a presentable technological justification in automation.
This reality doesn't change underlying trends, but it complicates the interpretation of statistics: when a company announces 5,000 job cuts "linked to AI," how many are truly due to automation and how many are related to a more traditional strategic review?
The question of redistribution is equally pressing: even if AI creates as many jobs as it globally destroys, these new jobs will not be created in the same places, in the same sectors, nor be accessible to the same profiles; a 52-year-old accountant in a provincial SME doesn't have the same retraining path as a 28-year-old developer in Paris.
The social challenge, therefore, lies not so much in the volume of jobs as in their geographical, social, and generational distribution.
No one truly knows what will happen, and any ten-year projection in such a rapidly evolving field must be taken with a grain of salt. What we can state with reasonable certainty is that jobs will change profoundly, some will disappear, others will emerge, and the speed of this transformation will likely be unprecedented in the modern history of work.
What's at stake now is less the question "Will AI make us unemployed?" than "How do we prepare to work with AI?" Young professionals have intuitively understood this; according to Indeed's 2026 data, Generation Z uses AI at work daily twice as much as Generation X.
Not to be replaced, but to work faster, better, and on higher-value tasks.
Perhaps that's the real lesson from all previous technological revolutions: those who know how to work with the machine are not the ones it replaces.
