Stanford Digital Economy Lab updated a paper on August 12, using high-frequency data from payroll service ADP covering millions of American workers to reveal the real changes in the labor market after the widespread adoption of generative AI. The report indicates that recent graduates and young professionals are experiencing the most direct impact from AI, rather than widespread job displacement across the country.

"First-tier" job collapse, seasoned workers remain unaffected

In high-impact professions such as software development and customer service, the employment rate of young people aged 22 to 25 lags behind that of fields less affected by AI by about 19%, while this gap was only 13% in 2025. However, experienced workers in the same positions have not faced the same shortage, with employment remaining stable or even increasing, highlighting a clear generational divide.

The reason for the concentration of the impact at the entry level lies in the fact that these jobs rely heavily on "codifiable knowledge"—standardized, documented content taught through textbooks and written procedures, which is clear-cut and easily participated in or replaced by AI. In contrast, jobs that rely on practical experience, mentorship, and real-world accumulation of "tacit knowledge" are more likely to be assisted by AI, rather than replaced by it.

The study also points out that higher education can buffer the impact of AI on employment. For the 22 to 25 age group, the employment rate for jobs that are easier to automate has declined more sharply, while the situation for jobs where AI serves only as an assistant is much more complex—offering a clear signal for education and career choices: holding onto abilities that are hard to codify is the anchor point for standing firm in the AI wave.