New study sheds light on what kinds of workers are losing jobs to AI

Artificial intelligence is replacing entry-level workers whose jobs can be performed by generative AI tools like ChatGPT, a rigorous new study finds.

Early-career employees in fields that are most exposed to AI have experienced a 13% drop in employment since 2022, compared to more experienced workers in the same fields and when measured against people in sectors less buffeted by the fast-emerging technology, according to a recent working paper from Stanford economists Erik Brynjolfsson, Bharat Chandar and Ruyu Chen.

The study adds to the growing body of research suggesting that the spread of generative AI in the workplace is likely to disrupt the job market, especially for younger workers, the report's authors said.

"These large language models are trained on books, articles and written material found on the internet and elsewhere," Brynjolfsson told CBS MoneyWatch. "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 with between these LLMs and the knowledge young people have."

The research highlights two fields in particular where AI already appears to be supplanting a significant number of young workers: software engineering and customer service. Between late 2022 and July 2025, entry-level employment in those areas declined by roughly 20%, according to the report, while employment for older workers in the same jobs grew.

Overall, employment for workers aged 22 to 25 in the most AI-exposed sectors dropped 6% during the study period. By comparison, employment in those areas rose between 6% and 9% for older workers, according to the researchers.

The analysis reveals a similar pattern playing out in the following fields:

• Accounting and auditing
• Secretarial and administrative work
• Computer programming
• Sales

Older employees, who generally have navigated the workplace for a longer period of time, are more likely to have picked up the kinds of communication and other "soft" skills that are harder to teach and that employers may be reluctant to replace with AI, the data suggests.

"Older workers have a lot of tacit knowledge because they learn tricks of trade from experience that may never be written down anywhere," Brynjolfsson explained. "They have knowledge that's not in the LLMs, so they're not being replaced as much by them."

The study is unusually robust given that generative AI technologies are only a few years old, while experts are just starting to systematically dig into the impact on the labor market. The Stanford researchers used data from ADP, which provides payroll processing services to employers with a combined 25 million workers, to track employment changes for full-time workers in occupations that are or more or less exposed to AI. The data included detailed information on workers, including their ages, and precise job titles.

AI doesn't just threaten to take jobs away from workers. As with past cycles of innovation, it will render some jobs extinct while creating others, Brynjolfsson said.

"Tech has always been destroying jobs and creating jobs. There has always been this turnover," he said. "There is a transition over time, and that's what we are seeing now."

For example, in fields like nursing AI is more likely to augment human workers by taking over rote tasks, freeing health care practitioners to spend more time focusing on patients, according to proponents of the technology.

While entry-level employment has fallen in professions that are most exposed to AI, no such such decline has occurred in jobs where employers are looking to use these tools to support and expand what employees do.

"Workers who are using these tools to augment their work are benefiting," Brynjolfsson said. "So there's a rearrangement of the kind of employment in the economy."

Workers who can learn to use AI to to help them do their jobs better will be best positioned for success in today's labor market, according to Brynolfsson.

A recent report from AI staffing firm Burtch Works found that starting salaries for entry-level AI workers rose by 12% from 2024 to 2025.

"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," Brynjolfsson told CBS MoneyWatch.

"This is the first time we're getting clearer evidence of these kinds of employment effects, but it's probably not the last time," he added. "It's something we need to pay increasing attention to as it evolves and companies learn to take advantage of things that are out there."