Introduction
AI’s impact on jobs is the focus of new Stanford research using high-frequency administrative data from the largest U.S. payroll software provider. The study surfaces six facts about how generative AI adoption is reshaping employment. Early-career workers (ages 22–25) in the most AI-exposed occupations see the sharpest declines, while more experienced peers and less-exposed fields remain stable or keep growing. Adjustments occur mainly through employment, not compensation, and drops are concentrated where AI is likely to automate rather than augment human labor. The results hold under alternative checks, such as excluding tech firms and remote-friendly roles.
In short: Stanford finds a 13–16% decline for ages 22–25 in the most AI-exposed jobs.
Context
Drawing on broad payroll coverage, the researchers show that aggregate averages hide unequal effects. Once roles and experience are separated, clear patterns emerge: software development and customer support—among other exposed functions—face measurable entry-level declines since late 2022, while senior workers in the same jobs do not. This suggests AI is already shifting the composition of employment where core tasks can be automated. Where AI augments human work, hiring is steadier. Robustness checks limit confounders by removing tech-heavy employers and remote-amenable occupations, indicating that the observed changes align with AI adoption rather than cyclical or policy noise.
Six key facts
These points summarize how AI currently affects junior workers across roles and organizational choices.
- Entry-level employment down: -13% (relative) for ages 22–25 in AI-exposed jobs
- External coverage reports a -16% drop since late 2022 for the same cohort
- Impact concentrates where AI automates, not where it augments
- Adjustment occurs via employment more than via wages
- Senior workers and less-exposed fields remain stable or grow
- Findings are robust after excluding tech firms and remote-friendly roles
Interpreting 13% vs 16%
Stanford reports a 13% relative decline for 22–25-year-olds in the most exposed occupations; media coverage on ADP data cites 16% since late 2022. Both point to a rapid, significant effect on entry-level hiring.
The Challenge
The core risk is a talent-pipeline bottleneck: if entry-level hiring shrinks, industries may lack experienced workers in a few years. This can slow innovation and widen generational gaps, especially where AI displaces codified or routine tasks. Benefits remain concentrated in a few categories (coding, customer service), while broad gains are harder to realize. The speed of change—likened to the pandemic-era shift to remote work—calls for updated training, apprenticeship-like pathways, and AI deployment strategies that balance productivity with human capital development.
"This is the fastest, broadest change that I've seen, comparable only to the shift to remote work during the pandemic."
Erik Brynjolfsson, Economist and Stanford Professor
Solution / Approach
Employment outcomes hinge on how firms deploy AI. Evidence suggests more hiring where AI augments people and less where it replaces them. A practical response includes: 1) redesigning jobs for augmentation-first workflows; 2) creating bridge roles for juniors in QA, model orchestration, and edge-case handling; 3) investing in mentorship to transfer tacit knowledge; 4) using near-real-time economic dashboards to align training with demand.
"Senior workers have tacit know-how and tradecraft that aren’t always written down—areas where AI does not excel, at least yet."
Erik Brynjolfsson, Economist and Stanford Professor
Conclusion
Combined signals (13–16% decline for ages 22–25 in exposed roles) indicate AI’s labor impact is arriving first at the point of entry. Because outcomes differ between automation and augmentation, design choices matter. The imperative is twofold: safeguard the talent pipeline and capture productivity gains. Continuous information tools and targeted HR policies can reduce the risk of a “missing cohort” in pivotal fields.
FAQ
Quick answers based on Stanford’s evidence in the U.S. labor market.
- How is AI’s impact on jobs affecting young workers? A 13–16% decline for ages 22–25 in the most AI-exposed roles, concentrated at entry.
- Where is the labor market impact strongest? In occupations where AI automates core tasks; augmentation correlates with steadier hiring.
- Are wages falling as much as jobs? Adjustments occur mainly through employment rather than pay.
- Does AI’s impact on jobs extend to senior workers? Thus far, experienced workers remain stable or grow; future risk is an open question.
- Are these findings robust to confounders? Yes; excluding tech firms and remote-friendly roles preserves the results.
- What can companies do now? Design for augmentation, create junior bridge roles, and strengthen mentorship and training.