Uneven Impact of AI on Employment by Age: Insights from Stanford Payroll Data
In November 2025, three Stanford economists—Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen—published a compelling study analyzing monthly payroll records from ADP, the largest payroll processor in the United States. Their research, titled “Canaries in the Coal Mine?”, leverages data covering over 25 million workers, with a focused sample tracking 3.5 to 5 million employees monthly from January 2021 through September 2025. This rich dataset reveals a nuanced story about AI’s impact on the labor market—one that challenges the broad narrative that “AI is coming for everyone’s job.”
Contrary to expectations, the impact of artificial intelligence on employment is not uniform across age groups. The study found that employment for 22-to-25-year-olds in the most AI-exposed occupations—such as software development, customer service, and clerical roles—fell roughly 13% after controlling for firm-specific shocks. In stark contrast, workers aged 35 to 49 in those identical roles experienced employment growth, with raw numbers showing an increase of over 8% during the same period.
These findings suggest that the immediate strain of AI-driven automation is disproportionately landing on younger workers. While the tools and job titles remain the same, the outcomes diverge sharply by age. Importantly, the paper notes that companies weren’t cutting salaries; annual salary trends showed little difference across age or exposure groups. Instead, the hiring slowdown focused on junior employees. For example, among software developers, the 22-to-25 age group saw nearly a 20% decline in employment from their late-2022 peak.
Understanding the Age Divide: Why Younger Workers Are More Affected
The authors propose an insightful explanation that upends common assumptions about AI and job automation. Typically, it might be expected that repetitive, experience-based tasks would be automated first, preserving opportunities for recent graduates with fresh credentials. However, the data suggests the opposite: AI models, trained extensively on written material such as books, articles, and documentation, effectively replicate the “book learning” that younger workers bring to their jobs.
As Brynjolfsson explains, “That’s the kind of book learning that a lot of people get at universities before they enter the job market, so there is a lot of overlap between these LLMs and the knowledge young people have.” In contrast, older workers possess significant tacit knowledge—unwritten tricks of the trade and experiential insights not captured in AI training data. This tacit knowledge acts as a buffer against automation, explaining why older employees are less affected by AI-driven displacement.
The study also distinguishes between AI’s use for automation versus augmentation. The employment declines among younger workers were concentrated in occupations where AI was automating tasks outright. Conversely, in roles where AI primarily augmented human work, junior employment remained stable or even grew. This nuance highlights that AI’s impact is multifaceted and context-dependent, rather than a monolithic force.
Supporting this pattern, Harvard economists Seyed Hosseini Maasoum and Guy Lichtinger found similar trends in their analysis of résumé and job-posting data covering roughly 62 million US workers across 285,000 firms. Their research showed that at firms adopting generative AI, junior employment dropped sharply relative to non-adopters starting in 2023, while senior employment continued to rise.
The Millennial Perspective: Navigating a Shifting Career Landscape
Reading these findings as a millennial offers a unique vantage point. While the 22-to-25 age group bears the brunt of AI-related job losses, older cohorts—including many millennials—cannot afford complacency. Career paths many have pursued are evolving faster than anticipated, with skill sets becoming obsolete more quickly than in previous decades. This reality means continuous adaptation is not optional but essential.
Brynjolfsson’s advice resonates here: “Young workers who learn how to use AI effectively can be much more productive. But if you are just doing things that AI can already do for you, you won’t have as much value-add.” This points to the critical importance of developing skills that complement AI, rather than compete with it.
Personal experience echoes this uncertainty. One author recounts leaving a relatively stable role running an adult language school to take a venture capital internship, a move that meant starting over in many ways but eventually proved worthwhile. The lesson is that a shaky foothold early in a career doesn’t preclude future success—it may simply require more resilience and adaptability.
Rethinking Career Strategies in the Age of AI
It’s important to approach these findings with a balanced perspective. Brynjolfsson reminds us that “tech has always been destroying jobs and creating jobs. There has always been this turnover.” History is full of automation fears that didn’t materialize as predicted. Whether AI will follow a similar trajectory or disrupt the labor market in fundamentally new ways remains to be seen.
What the payroll data does underscore is that early career stages—the entry rungs of the ladder—are the most vulnerable to disruption. The assumption that credentials and entry-level positions guarantee stability is increasingly tenuous. Instead, the landscape calls for proactive engagement with AI tools and an emphasis on uniquely human skills that AI cannot easily replicate.
Ultimately, the story is still unfolding. The current data could represent an initial shock that ripples upward as AI systems improve at capturing tacit knowledge, or it might be a more localized adjustment confined to junior roles. More research and time will clarify which narrative holds true. Until then, the prudent path for workers and employers alike is vigilance, adaptability, and a commitment to lifelong learning.
For the full study and detailed insights, see the source Here.
