AI exposure is reshaping job exits among older US workers

Older US workers in occupations more exposed to artificial intelligence are leaving jobs at higher rates, according to new research from the Center for Retirement Research at Boston College. The finding matters for employers and policymakers because longer working lives are a central assumption in debates over Social Security solvency and retirement-age reform.

Geoffrey Sanzenbacher, an economics professor and the paper’s author, told CNBC that workers aged 55 and above in AI-exposed industries are moving out of jobs more frequently, with exits split between unemployment and voluntary departures. He described the effect as statistically significant and said it can be substantial in some occupations.

How the research measures exposure

The paper defines AI exposure by assessing how far artificial intelligence can be used to perform tasks within a given occupation. It draws on Current Population Survey data and AI-exposure measures from Tufts University’s Digital Planet initiative, which studies the economic and social effects of digital technologies.

Sanzenbacher’s research identifies three possible channels. Automation can replace some older workers, leaving them unemployed or out of the labor force. Some employees may also leave roles that require new AI tools, either for jobs less affected by the technology or for retirement. A third possibility is that generative AI increases productivity, supports higher wages and enables some workers to remain employed for longer by shifting time toward tasks they find more engaging.

Before OpenAI released ChatGPT, older workers in roles with high AI exposure were significantly less likely to leave their jobs, according to the research. After ChatGPT’s launch, they became somewhat more likely to move out of work, including into unemployment.

White-collar roles face the highest exposure

The Boston College paper finds that older workers most exposed to AI tend to be white, more likely to hold college degrees and higher-earning than workers in occupations with low AI exposure.

Based on Digital Planet’s AI-exposure scores, the five occupations with the highest exposure are web and digital interface designers, web developers, database architects, computer programmers and data scientists. The lowest-exposure occupations listed in the research are excavating and loading operations and mining workers, roof bolters and mining workers, orderlies, painting and spraying workers, and fiberglass laminators and fabricators.

The pattern challenges a standard assumption in retirement policy: that workers in physically demanding, lower-paid jobs are more likely to have shorter careers than higher-paid white-collar employees. “AI exposure may reduce the gap in career length between low- and high-paying jobs,” Sanzenbacher wrote.

Social Security debate adds urgency

The latest annual report from Social Security’s trustees projects that the trust fund used to help pay retirement benefits could be depleted in late 2032. Policymakers seeking to restore solvency could consider measures including higher retirement ages or higher payroll taxes on high earners. The last major Social Security overhaul, enacted in 1983, gradually raised the retirement age from 65 to 67.

Sanzenbacher told CNBC that higher-income people have a high probability of seeing larger benefit reductions than lower-income people under future Social Security changes. He said those workers may therefore need to extend their careers, even as AI could affect their ability to remain in current roles.

Adoption remains uneven

AARP research found that workers aged 50 and older view AI as both risk and opportunity. In a March survey of 1,015 adults in that age group, 24% said AI was a threat to their line of work, 19% called it an opportunity and 37% said it was both.

Separate research from AARP and LinkedIn found that experienced professionals are more likely than younger workers to hold roles insulated from generative AI disruption, at 49.4% versus 42.2%. The research said older workers’ jobs more often require skills AI cannot easily replicate, including collaboration, judgment and leadership.

Monster’s December WorkWatch report found that 42% of 1,504 workers surveyed do not use AI at all. Among users, Monster said common applications include email, scheduling and writing support, while some workers use AI for coding, automation, data analysis, job applications and creative work.

Vicki Salemi, a career expert at Monster, told CNBC by email that older workers can still begin building AI literacy. She said workers may start with tools already used by their employers while continuing to emphasize communication, relationship-building and problem-solving skills.

This story draws on original reporting from CNBC.