Experience matters: Decoding first signs of AI’s impact on job market

The study is based on millions of records from ADP, America’s largest payroll processing company

A new study leveraging massive real-time US payroll data has found that generative AI has begun having significant and disproportionate impact on entry-level workers in certain jobs, including software developers and customer service agents.

The findings carry particular significance for countries with large young populations entering AI-exposed sectors. (Representational image)

The study by researchers from Stanford University’s Digital Economy Lab, noted workers in AI-exposed occupations experienced a 13% decline in employment between late 2022 and July 2025 while their colleagues over 35 in identical roles maintained stable or growing job prospects. For these older workers, job opportunities grew 6-9%.

“For highly AI-exposed jobs like software developers and customer service agents, people aged 22 to 25 saw a very striking decline in employment in the last few years in the US,” said Bharat Chandar, one of the Stanford University researchers who authored the paper, in an interaction over email .

When determining which jobs are AI exposed, the research zeroed in on an important distinction: can these AI tools automate these jobs, or do they work to augment an employee’s work?

Chandar said the disruptions are more evident in occupations where LLMs (large language models– the type of technology behind ChatGPT, or Claude) “automate work than for ones where it augments work”. For instance, “for less-exposed jobs such as health aides we do not see these declines”.

The insight might be intuitive, but it is now backed by empirical evidence – at least from the US market. The study – titled Canaries in the Coal Mine – is based on millions of records from ADP, America’s largest payroll processing company, allowing researchers to track employment and earning changes with granularity typically not possible with official labour statistics.

The analysis found that employment for software developers aged 22-25, for instance, fell nearly 20% from its peak in late 2022, while workers over 35 in the same field saw employment grow by 6-9%. Customer service representatives, another AI-automated role, showed similar age-based divergences.

Chandar added that while the automation-augmentation is “not a perfect causal” factor, it withstood comparisons with other possible causes. “We tested various alternative hypotheses such as remote work, interest rate changes, and tech over-hiring that we thought might explain the results and found the same clear pattern between AI exposure and employment even when accounting for these other factors”.

In addition to computer programmers, other roles such as financial managers, accountants and sales representatives -- all occupations where AI can directly automate core tasks -- showed similar employment declines for young workers.

On the other hand, maintenance and repair workers, registered nurses, and computer systems managers, maintained stable or growing employment across age groups.

Crucially, the employment effects appeared exclusively in hiring rather than compensation, with wages remaining largely unchanged across age groups and exposure levels -- suggesting companies are adjusting workforce composition rather than cutting pay.

The findings carry particular significance for countries with large young populations entering AI-exposed sectors. India’s information technology industry, which employs millions in software development and business process outsourcing, may face similar dynamics.

“India is a particularly interesting case to study given the number of young people who work in service jobs that may compete with AI,” Chandar said, though he cautioned that the researchers did not want to extrapolate too much from US data.

About a month ago, India’s largest IT company Tata Consultancy Services announced plans to cut 12,000 jobs, and this week formed a new AI and services transformation unit — illustrating the trade-offs the Stanford study identified between AI investment and human workforce.

“I do think more attention should be directed towards the labour market changes in places like India where a substantial number of people work in occupations that may be exposed to AI,” Chandar said. “We hope to track this ourselves in the near future and encourage others to study this as well.”

India produces over 1.5 million engineering graduates annually, many entering fields the study identifies as highly exposed to AI automation, potentially magnifying the employment effects documented in American data.

The researchers offered a theoretical explanation for why AI disproportionately affects younger workers.

Codified versus tacit knowledge

The researchers proposed that AI’s disproportionate impact on young workers stems from fundamental differences in the types of knowledge they possess compared to experienced colleagues.

“One hypothesis for why we see these impacts more for young workers than for older workers is because they have more codified knowledge, such as what students learn from books or in the classroom,” Chandar explained.

This creates a vulnerability for recent graduates. “LLMs may be trained and developed using similar information found on the internet, which could result in overlap between LLM capabilities and the sort of tasks that young workers have the most ability to perform,” he said.

In contrast, “older workers may rely more on tacit knowledge, which is the tips and tricks of the trade that aren’t written down but are learned via experience,” Chandar said.

This tacit knowledge includes institutional memory, relationship skills, and problem-solving approaches developed through years of practice -- capabilities that resist easy replication by AI.

But there are also some outliers. Employment effects of AI vary by education level. “Less educated workers in exposed jobs see employment declines even at higher ages,” Chandar noted, suggesting that for workers in occupations requiring fewer qualifications, the protective value of experience accumulates more slowly.

Monitoring required

The researchers emphasised the preliminary nature of their findings, noting that technological disruptions historically create adjustment periods where some workers benefit more than others. Over time, workers typically shift from displaced roles to emerging areas of demand. “We think it is very important to study these questions in other countries as well, and we hope we can do that in the coming weeks and months,” Chandar said.

The Stanford team plans to continue tracking employment patterns to determine whether these represent temporary adjustment effects or permanent structural changes in the labour market, providing guidance for supporting workers who may be affected by AI-driven disruption.