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

More U.S. managers are utilizing AI, survey shows

AI could find if you'll get a raise, promotion oregon look layoffs 03:39

Artificial quality is replacing entry-level workers whose jobs tin beryllium performed by generative AI tools similar ChatGPT, a rigorous caller survey finds.  

Early-career employees successful fields that are astir exposed to AI person experienced a 13% driblet successful employment since 2022, compared to much experienced workers successful the aforesaid fields and erstwhile measured against radical successful sectors little buffeted by the fast-emerging technology, according to a recent working paper from Stanford economists Erik Brynjolfsson, Bharat Chandar and Ruyu Chen.

The survey adds to the growing assemblage of research suggesting that the dispersed of generative AI successful the workplace is apt to disrupt the occupation market, particularly for younger workers, the report's authors said. 

"These ample connection models are trained connected books, articles and written worldly recovered connected the net and elsewhere," Brynjolfsson told CBS MoneyWatch. "That's the benignant of publication learning that a batch of radical get astatine universities earlier they participate the occupation market, truthful determination is simply a batch of overlap with betwixt these LLMs and the cognition young radical have." 

The probe highlights 2 fields successful peculiar wherever AI already appears to beryllium supplanting a important fig of young workers: bundle engineering and lawsuit service. Between precocious 2022 and July 2025, entry-level employment successful those areas declined by astir 20%, according to the report, portion employment for older workers successful the aforesaid jobs grew. 

Overall, employment for workers aged 22 to 25 successful the astir AI-exposed sectors dropped 6% during the survey period. By comparison, employment successful those areas roseate betwixt 6% and 9% for older workers, according to the researchers. 

The investigation reveals a akin signifier playing retired successful the pursuing fields:

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

Older employees, who mostly person navigated the workplace for a longer play of time, are much apt to person picked up the kinds of connection and different "soft" skills that are harder to thatch and that employers whitethorn beryllium reluctant to regenerate with AI, the information suggests. 

"Older workers person a batch of tacit cognition due to the fact that they larn tricks of commercialized from acquisition that whitethorn ne'er beryllium written down anywhere," Brynjolfsson explained. "They person cognition that's not successful the LLMs, truthful they're not being replaced arsenic overmuch by them."

The survey is unusually robust fixed that generative AI technologies are lone a fewer years old, portion experts are conscionable starting to systematically excavation into the interaction connected the labour market. The Stanford researchers utilized information from ADP, which provides payroll processing services to employers with a combined 25 cardinal workers, to way employment changes for full-time workers successful occupations that are oregon much oregon little exposed to AI. The information included elaborate accusation connected workers, including their ages, and precise occupation titles.  

AI doesn't conscionable endanger to instrumentality jobs distant from workers. As with past cycles of innovation, it volition render immoderate jobs extinct portion creating others, Brynjolfsson said. 

"Tech has ever been destroying jobs and creating jobs. There has ever been this turnover," helium said. "There is simply a modulation implicit time, and that's what we are seeing now."

Augmented oregon automated?

For example, successful fields similar nursing AI is much apt to augment quality workers by taking implicit rote tasks, freeing wellness attraction practitioners to walk much clip focusing connected patients, according to proponents of the technology. 

While entry-level employment has fallen successful professions that are astir exposed to AI, nary specified such diminution has occurred successful jobs wherever employers are looking to usage these tools to enactment and grow what employees do. 

"Workers who are utilizing these tools to augment their enactment are benefiting," Brynjolfsson said. "So there's a rearrangement of the benignant of employment successful the economy." 

Advice for young workers

Workers who tin larn to usage AI to to assistance them bash their jobs amended volition beryllium champion positioned for occurrence successful today's labour market, according to Brynolfsson. 

A caller report from AI staffing steadfast Burtch Works recovered that starting salaries for entry-level AI workers roseate by 12% from 2024 to 2025. 

"Young workers who larn however to usage AI efficaciously tin beryllium overmuch much productive. But if you are conscionable doing things that AI tin already bash for you, you won't person arsenic overmuch value-add," Brynjolfsson told CBS MoneyWatch. 

"This is the archetypal clip we're getting clearer grounds of these kinds of employment effects, but it's astir apt not the past time," helium added. "It's thing we request to wage expanding attraction to arsenic it evolves and companies larn to instrumentality vantage of things that are retired there." 

• Artificial Intelligence