According to ADP data, employment for those aged 22 to 25 has decreased by 13%. In roles heavily affected by automation, there has been a steep decline, whereas roles enhanced by AI have remained stable. “Without mentoring and on-the-job training, the skilled workforce pipeline may collapse.”
“Is artificial intelligence (AI) taking away jobs from young workers?” A new report released by Stanford University’s Digital Economy Lab provides a concrete basis for this question. The researchers used the metaphor of the “canary in a coal mine” to warn that the potential impact of AI on the labor market can already be observed in the employment reduction of the early-career age group (22 to 25 years old).
The study utilized administrative data from ADP, the largest payroll management company in the United States. This data, comprising monthly employment and wage records for millions of employees and tens of thousands of companies, is considered one of the largest and most real-time analyses to date.
The study found that in AI-exposed occupations (such as software engineering, marketing, and customer support), employment for those aged 22 to 25 decreased by about 13% from the end of 2022 to July 2025. This contrasts with a 6-12% increase in employment for workers over 30 in the same occupations. Erik Brynjolfsson, a Stanford professor and one of the authors of the study, stated that overall employment continues to rise, but young workers are facing unique challenges, which is a clear signal.
The researchers analyzed conversation logs from Anthropic’s Claude to distinguish whether AI use was centered on automation or augmentation in various roles. They found that significant reductions in early-career employment were concentrated in roles with a strong automation focus, while jobs with a strong emphasis on augmentation showed little employment decline. This indicates that how AI is deployed can determine the impact on employment. Professor Brynjolfsson emphasized that while technology has always displaced jobs, it also creates new ones. He suggested that AI, if used in an augmentative way, could lead to not only increased productivity but also inclusive prosperity.
This study provides statistical evidence for the hiring struggles recent college graduates in the U.S. are facing in fields like software and customer service. The researchers mentioned that the study aligns with the hypothesis that AI is affecting the labor market, independent of factors like COVID-19, remote work expansion, or changes in the education system.
The researchers called for policy responses, as losing opportunities for early-career workers to gain hands-on experience could potentially disrupt the skilled workforce pipeline in the long term. They suggested strengthening mentoring, on-the-job training, and education on augmentative AI use.
Interestingly, the researchers also used AI during the drafting process of their paper. AI contributed to writing code for data cleaning, cross-referencing citations, creating and editing graphs, and drafting some documents. The researchers explained that they could complete the paper efficiently and effectively by using AI as an augmentative tool. [Silicon Valley Correspondent Won Ho-seop]