Brief IA
›

AI: Recruitment Facing the Risk of Expertise Dwindling

Business & Startups·Tom Levy·

AI: Recruitment Facing the Risk of Expertise Dwindling

AI: Recruitment Facing the Risk of Expertise Dwindling
Key Takeaways
1Researchers from Stanford measure a relative decline of 13% in employment for 22-25 year-olds in jobs most exposed to AI.
2A survey conducted with France Travail in early 2025 indicates that 35% of job seekers have used AI for their CVs.
3Anthropic's extreme scenario predicts a share of wealth allocated to employees dropping from 60% to 45%.
Why it matters : Jean Eudes Yahouedeou fears that the automation of entry-level tasks will deprive companies of their future talent.

Youth employment is declining in American jobs heavily exposed to AI, according to researchers from Stanford. In recruitment, where AI-assisted applications are multiplying, Jean Eudes Yahouedeou, founder and CEO of Seeqle, advocates for a learning approach centered on machine control and decision-making responsibility.

Recruiters Must Also Assess Judgment

According to a survey conducted with France Travail in early 2025, 35% of job seekers have already used AI to write their CVs. Half of companies also report an increase in machine-generated applications.

Jean Eudes Yahouedeou, who markets AI to HR services, argues that eliminating entry-level tasks risks depriving companies of their future experts, while judgment and personal responsibility would gain in value. He describes recruitment as being a few years ahead of this evolution and believes that a CV that has become uniformly flawless loses its evidentiary power. His proposal is to select beginners based on their verification, arbitration, and accountability skills; by 2030, their training would involve machine supervision alongside an experienced employee, a skill that a CV would not be able to identify.

In his team, he reports, AI initially took over tasks previously reserved for interns: extraction and sorting, synthesis, transposition and transformation, analysis, testing, or follow-ups. He distinguishes between manual knowledge and experience and sees in the sorting of applications, contractual verification, or IT correction the training of the experts of 2035. His scenario is conditional: if companies automate these activities and forgo hiring beginners, the savings and productivity could mask a shortage of replacements as seniors retire for years. He compares this delay to that of a forest where young trees have disappeared, with no apparent effect for twenty years.

To measure the value of work, the entrepreneur therefore proposes signed decisions, rather than hours logged by Keynes or tasks counted by Anthropic. His examples involve a candidate hesitant to convince, an arbitration between two imperfect applications, or a choice to justify before management: in case of failure, he reminds, the recruiter must be accountable, not the algorithm. He anticipates machines becoming increasingly cheaper for activities that do not involve anyone, a greater value for those that entail responsibility, and thus defends the maintenance of learning.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Stanford Observes a Gap Between Youth and Seniors

In August 2025, researchers from Stanford University examined payroll data from millions of people employed in the United States. Their findings focus on those aged 22-25: since the emergence of ChatGPT, their employment has decreased by 13% in the professions most exposed to AI, compared to other professions.

More experienced employees in these same jobs do not show this effect, according to the team. The result thus distinguishes situations based on experience within the affected professions.

Anthropic Projects a Smaller Share for Employees

Anthropic's economic projections concern the United States in 2030. The developer of Claude presented three hypotheses on September 9, all linking economic expansion to a share of income devoted to work that is increasingly lower as AI progresses. The most moderate scenario envisions a change comparable to that caused by the Internet.

In the intermediate hypothesis, AI could perform half of intellectual tasks and growth would double. However, salaries in intellectual professions would stop progressing, unlike those in other jobs. The extreme hypothesis anticipates a one-third increase in the economy, but a reduction in the share of wealth allocated to employees from 60% to 45%, with a decrease of over 10% in intellectual salaries.

Anthropic sells the evaluated technology and is preparing for an IPO. Its model has been reviewed by David Autor and Daron Acemoglu, the 2024 Nobel laureate in economics; some reviewers even find the hypothesis cautious compared to already observed phenomena.

The historical precedent cited by Yahouedeou dates back to 1930: in a letter to his grandchildren, John Maynard Keynes imagined a wealth four to eight times greater by 2030 and fifteen hours of work per week. The entrepreneur considers the first prediction to be accurate, but not the second: in his view, the time freed up by machines has been devoted to more production and consumption.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.