He’s done it again. A few days ago on the American TV channel Fox News, Dario Amodei, CEO of Anthropic, the parent company of the generative AI model Claude, claimed that some 50% of all technology jobs, entry-level lawyers, consultants and finance professionals will be eliminated within one to five years.
“Entry-level jobs will be replaced by AI systems,” he told Fox. “We may indeed have a serious employment crisis on our hands.”
The entrepreneur had already expressed similar views several times; each time, his declaration causes a sensation. Especially since everyone senses that AI is an innovation unlike any other.
In France, for now, in most companies, employees themselves are adopting these new tools without waiting for approval from their organizations. Everyone experiments privately, and quietly reorganizes their workdays around productivity gains they discover on their own. Thanks to Claude itself, presentations that once required several days of work are now completed in an hour. Information monitoring is automated. Research is done in the blink of an eye.
Most HR departments are well aware of these innovations but still hesitate to draw all the consequences in terms of organizational restructuring. At most, some companies have begun announcing—internally at least—that they will take on significantly fewer interns this summer.
But what about two years from now? Employees in major service companies will all go on vacation this summer with that question in mind. Since the invention of the weaver’s loom, the end of work has been predicted a thousand times. And each time, work reinvented itself. But what if this time is different? Could capitalism’s cycle of creative destruction continue without humans? The uncertainty surrounding the pace and manner of AI adoption by companies fuels these apocalyptic scenarios.
One of Dario Amodei’s peers, researcher Yann LeCun, former chief AI scientist at Meta, has offered another perspective. “Dario is wrong. He knows absolutely nothing about the effects of technological revolutions on the labor market. Don’t listen to him, Sam, Yoshua, Geoff, or me on this topic,” he wrote on X, referring to other AI experts. “(Instead) listen to economists who have spent their career studying this.”
The findings of their research resonate less loudly in public debate because their models, subject to numerous uncertainties, cannot be reduced to slogans. One of the most discussed recent papers comes from Americans Daron Acemoglu, David Autor and Simon Johnson. To understand AI’s future effect on the world of work, the MIT economists distinguish five different technologies: those that augment labor, those that augment capital, those that automate, those that level out expertise, and finally those that create new tasks. The idea, of course, is to encourage companies to adopt applications favorable to workers.
A 5% increase in employment
For now, and contrary to common narratives, the first figures regarding AI adoption in companies are reassuring. Businesses that deployed AI solutions from their earliest years did not use them to cut their workforce.
A study by Philippe Aghion and co-authors, based on 2023 data from France’s INSEE, shows that after adopting AI, French companies tended to increase both employment and revenue. This conclusion suggests that productivity gains offered by AI allow companies to grow, compensating—at least in part—for employment substitution effects linked to automation.
On average, according to their calculations, sales and employment within these companies increased by about 5%. More surprisingly, the researchers found positive effects on staffing levels even in professions considered most vulnerable to automation.
For now, then, the classic engines of growth remain resilient. Far from destroying entire sectors of the economy and sending millions of workers home, AI is generating growth.
“If tomorrow companies make more money with fewer employees thanks to AI, that money will ultimately be reinjected into the economy in one way or another, through consumption or investment,” economist Augustin Landier explained a few weeks ago in L’Express.
What remains unclear is what forms of employment will be created through this new consumption or investment—and whether workers will benefit from them.
“Technology doesn’t do the housework”
Everything still appears wide open. Traditionally, automation movements do not eliminate professions but rather certain tasks that previously defined a role. They trigger a redefinition of jobs.
Employment in France’s retail sector, for example, has remained stable for 20 years despite checkout automation. Tasks have evolved: former cashiers supervise several automated stations—which is not necessarily less stressful, since interactions with customers become less predictable—while others advise shoppers, prepare orders for logistics services, and so on.
In banking, employment is eroding, but at a much more gradual pace than the shift toward app-based customer relations might suggest. Jobs have emerged in online advisory, predictive credit modeling, regulatory compliance and ecological transition.
It will be up to companies in the coming months to carefully choose which new tasks they automate with AI. Their challenge will be transforming employees’ jobs without sacrificing motivation in pursuit of short-term profitability.
Economist Erik Brynjolfsson recently measured the dangerous gap between workers’ wishes and the plans of AI experts, who often propose delegating to software tasks that appear repetitive and simple but are actually appreciated—and even viewed as central—by employees.
“We want technology to do the dishes and household chores while we write poems. But not only does current technology still not do the housework, it writes poems for us instead!” economist Antonin Bergeaud mischievously explains in his recent essay Recovered Prosperity (La Prospérité retrouvée, Odile Jacob, 2026).
Young people on the front line
For now, it is mainly young graduates who are bearing the brunt of the first waves of AI deployment.
In the United States, a study published at the end of 2025 using high-frequency payroll administrative data highlighted a relative employment decline of around 16% among 22- to 25-year-olds in the professions most exposed to AI, from the spread of generative AI through September 2025.
In its latest economic report, INSEE also notes that in France, in information technology and information services sectors, “since 2023, employment has declined while value added has continued along a trajectory without major change.”
The institute notes that the phenomenon primarily affects young people entering the labor market.
“More specifically, AI-related adjustments may in the short term concentrate less on total employment than on hiring structures, particularly entry-level positions in certain professions (support functions, administration, consulting, certain development and analysis tasks), and strongly penalize young people,” the experts write.
One of the challenges facing HR departments in the coming years will therefore be designing new pathways for young recruits.
Several approaches are already under debate: preserving simple, codified and repetitive tasks for training purposes, or on the contrary using AI to allow new hires direct access to more complex assignments without first going through step-by-step learning of codified tasks.
It is a daunting challenge that will require intense human support as well as a revolution in higher education.
“Experience would no longer be built through gradual accumulation of gestures and routines, but by plunging immediately into situations requiring reflection, decision-making and responsibility,” Bergeaud writes.
Competition and inequality
Beyond questions of employment levels, economists are divided over AI’s impact on inequality.
Will the new technology, by giving everyone access to tools previously reserved for a small elite, produce an equalizing effect among workers?
Far from that rosy scenario, current stock market valuations of major tech companies instead point toward a concentration of wealth among capital holders.
“The AI revolution, like all revolutions since the 19th century, will generate productivity gains and a long-term reduction in working time,” says executive and essayist François-Xavier Oliveau.” In the 19th century, work occupied between 40% and 50% of our waking lives. Today, it accounts for 12%. That proportion will continue to decline. The effect of this revolution is fundamentally deflationary. Goods and services will become cheaper. But labor income also risks coming under heavy pressure. There will be a real issue concerning the distribution of wealth increasingly created by capital.”
Closely tied to inequality is the question of competition. Until a few months ago, the general assumption was that AI would favor the wealthiest companies, since they could adopt new software more quickly.
The deployment of Claude Code, an agentic coding tool that has gone viral since the beginning of the year, has disrupted that perception.
Four experts recently noted in a paper on the Telos website: “The near-monopolies of major SaaS players such as business software specialist SAP could face competition from new players capable of offering customized products to companies: the economies of scale that allowed a company of 100,000 people to dominate a global market could disappear.”
Until the next twist...
