Employment of developers aged 22 to 25 has fallen by nearly 20% since the end of 2022, while AI now represents 26.9% of code in production and nearly a third of code merged by daily AI users. This divide reveals a paradox: artificial intelligence spares young people from mass unemployment, but disrupts their career trajectories.
AI is reshaping the value of human work according to logics that favor tacit experience over the codified knowledge of recent graduates. This transformation challenges traditional models of education and professional integration in the American economy.
Key Points
- Workers aged 22-25 in roles most exposed to AI are experiencing a 13% employment decline since late 2022
- Teams using AI assistants gain 10 to 15% productivity improvements, but review times increase by 91%
- Workers with AI proficiency earn a 56% salary premium in 2024, compared to 25% the previous year
- 62% of organizations are experimenting with AI agents that automate junior employees’ traditional learning tasks
Young Graduates Facing Automation of Their Learning Curve
The decline in youth employment in AI-exposed sectors results primarily from a drop in direct transitions from inactivity to employment rather than from layoffs. Companies are no longer hiring for entry-level positions that once served as paid training.
Junior employees traditionally absorbed “tacit knowledge”—unwritten rules, cultural context, and complex judgment—by completing routine tasks. In 2025, AI agents have captured the domain of “codified knowledge.” AI now masters code generation and financial modeling, the very tasks juniors used to learn from.
70% of managers believe AI can perform intern-level work, and 57% trust AI output more than that of recent graduates. This perception transforms the economic equation of junior hiring.
Productivity Gains, But Not Where Expected
AI’s productivity gains create paradoxes. When experienced developers use AI tools, they spend 19% more time completing their tasks, even though they perceive themselves as 20% faster. This misperception of their efficiency reveals a major cognitive bias.
Teams using AI assistants see productivity gains of 10 to 15%, but often this saved time is not redirected toward higher-value work. These modest gains therefore fail to translate into positive returns.
Developers on teams with high AI adoption complete 21% more tasks and merge 98% more pull requests, but review time increases by 91%. Human bottlenecks quickly neutralize AI’s benefits.
Seniors Resist Technological Substitution Better
Access to AI tools like Copilot increases output by 26% on average. But the largest productivity gains concern recent hires and junior developers (27% to 39% improvement), while seniors progress by 8% to 13%.
This difference masks a counterintuitive phenomenon: employment for less-exposed or more experienced workers remains stable or even increases. In sectors most exposed to AI, employment has grown 6% to 9% for older workers.
Longer-tenured employees, who have navigated the labor market longer, are more likely to have acquired communication skills and other “soft skills” that are harder to teach and which employers are reluctant to replace with AI.
An AI Skills Premium That Deepens Inequality
Workers with AI skills earn an average salary premium of 56% in 2024, double the 25% from the previous year. PwC confirms that workers with AI proficiency earn a 56% salary premium compared to their peers in the same profession. Oxford University quantifies that Machine Learning skills can increase salaries by 40%.
Yet this premium concerns a limited segment. On the other hand, many remote workers in Europe and the United States performing AI training tasks find their effective hourly rate falling below the national minimum wage—sometimes as low as 4 to 7 dollars per hour.
According to the federal report on labor market outcomes, computer science graduates show one of the highest unemployment rates across all fields. With 7.5% unemployment, fine arts degree holders are more employed than software engineers. Computer science graduates also face a 6.1% unemployment rate, nearly a full percentage point higher than liberal arts graduates.
The Collapse of the Traditional Learning Model
Internships are declining 11% across all industries according to Indeed. Handshake reports a 30% drop in tech-specific internship offers since 2023. Meanwhile, internship applications have increased 7%.
A critical nuance of the 2025 labor market is the emergence of a “low hiring, low layoffs” equilibrium. Companies are achieving workforce reductions through attrition rather than mass layoffs. This dynamic harms entry-level candidates more than anyone. In a “low hiring” environment, the few available positions are prioritized for immediate impact. Organizations prefer to hire one experienced senior employee rather than three juniors requiring training.
Job postings requiring generative AI skills in non-tech roles have increased ninefold between 2022 and 2024, reaching over 29,000. But access to these opportunities requires skills that traditional curricula have not yet systematically provided.
Toward a New Generational Contract
The Future of Jobs Report 2025 projects 78 million net new roles by 2030, even as 22% of current jobs undergo structural change. Most employers plan to train their workforce, with 85% offering reskilling and 77% offering AI training, but 63% cite skill gaps as their main obstacle.
China Launches the World’s Largest Educational Experiment with 280 Million Students Trained in Artificial Intelligence, revealing the scale of necessary educational transformations. This initiative contrasts with the fragmented approach in the United States, where only 2.3% of Indian workers have formal skills training, compared to 75% in Germany.
Despite a 29% drop in entry-level openings year-over-year, many are diversifying their income streams and using AI at work. Young people are adapting their career strategies to a market that no longer offers the same traditional entry points.
The current transformation of the American labor market does not replicate past cycles of technological innovation. AI creates an unprecedented generational rupture where human experience acquired through work better resists automation than recent theoretical knowledge. This inversion of traditional employability logics redefines the value of human capital and questions educational models inherited from the twentieth century.
Sources:
- https://www.technologyreview.com/2026/04/13/1135675/want-to-understand-the-current-state-of-ai-check-out-these-charts/