admin
A 16% Drop in Jobs for Young Workers
Payroll data from ADP, the largest US provider, was analyzed by Stanford researchers Erik Brynjolfsson, Ruyu Chen, and Bharat Chandar from late 2022 (when ChatGPT launched) to mid-2025. They focused on customer service and software development, two sectors that have long been identified as vulnerable to generative AI. The results were striking: employment among workers aged 22 to 25 in these sectors fell by 16%.
However, not everyone shared that pain. More experienced employees in the same fields saw their opportunities hold steady, or even grow.
The researchers conclude that AI is displacing repetitive entry-level tasks and accelerating and expanding the capabilities of experienced workers. Methodology Matters
This wasn’t a casual glance at the job market. The Stanford group put a lot of effort into normalizing the data and controlling for what they call “shock events” that could cause the picture to be distorted. They thought about the pandemic’s lingering effects, the rise of hybrid work, and ongoing tech layoffs. Even after accounting for these factors, the AI signal remained clear.
“Employment has begun to decline for young workers in highly exposed occupations like coding and call centers,” Brynjolfsson summarized in a tweet. “But older workers and workers who use AI to augment, not automate work have seen job gains.”
Beyond Displacement: Policy and Preparation
So what should be done? Brynjolfsson has proposed modifying tax policy to prevent businesses from being compensated for replacing human workers with machines. He also advocates for new benchmarks that measure how well AI collaborates with humans, what he calls “centaur” systems of humans and machines working together.
Others see demand growing for roles that involve managing and refining AI output. Meaning the future may be less about wholesale replacement and more about learning how to work with AI effectively.
Nevertheless, Brynjolfsson warns that the disruption may extend beyond entry-level positions. He stated to WIRED, “What we need to do is create a dashboard early-warning system to assist us in tracking this in real time.” “This is a very consequential technology.”
