In August, Stanford economists Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen updated their study of how AI is affecting jobs. It draws on anonymized ADP payroll records covering 3.5 million to 5 million workers a month, now through June 2026.
The headline: employment of workers aged 22 to 25 in the most AI-exposed occupations is 19 percent below where it would be if it had kept pace with their peers in less-exposed jobs. Experienced workers in the same occupations show no comparable gap. Employment for workers aged 35 to 40 grew by as much as 11 percent.
Last year’s version of the study was widely reported as a 13 percent decline. That figure came from a different statistical measure, so the two numbers aren’t directly comparable. On the authors’ new, simpler measure, the gap was 15 percent on July 2025 data and is 19 percent now. They report that it has widened steadily since August 2025.
- There is no sign of economy-wide job loss. Average employment rose about 6 percent from November 2022 to June 2026.
- The gap comes mainly from less hiring, not more layoffs or departures.
- It is concentrated where AI is used to automate tasks rather than to assist people. For older workers, assistive use is associated with growth.
- The adjustment shows up in employment, not base pay.
- Software development and customer service are the case studies.
The authors call these descriptive facts, not proof of cause. Controlling for the share of college graduates cuts the estimate roughly in half. The payroll data overrepresents large firms and exposed occupations, and the gap is smaller in Census survey data, though it points the same way.
AI isn’t emptying offices. It’s narrowing the door new hires come in through.
Why this is a workforce planning problem
Today’s 22-year-old analysts and support staff are where tomorrow’s experienced staff come from. When hiring slows in roles where AI automates entry-level tasks, the work that new hires learned on goes with it. Nobody gets laid off, so the change is easy to miss on a headcount report.
Questions for workforce and L&D leaders
- Which of your entry-level roles overlap most with tasks AI now automates?
- Has hiring into those roles slowed, even without any layoffs?
- What did new hires learn from the work AI now does, and where will they learn it instead?
- Are you using AI to replace junior tasks, or to let junior staff take on harder work sooner?
- If the entry pipeline shrinks, who will be your experienced staff in five years?
Nobody was laid off. Fewer people were hired. The bench gets shorter either way.
Sources