AI Exposure Linked to Earlier Job Exits for Workers 55+

Workers approaching retirement in artificial intelligence-exposed occupations are exiting the workforce at elevated rates compared to the pre-ChatGPT era, according to new research from Boston College's Center for Retirement Research.

The analysis examined employment patterns before and after November 2022, when ChatGPT's public launch accelerated AI adoption across industries. Researchers found that older workers in roles with high AI exposure—including programmers, accountants, and data analysts—showed measurably higher job exit rates in the post-ChatGPT period.

Why it matters

As policymakers increasingly emphasize the need for extended working lives to ensure retirement security, AI adoption may be pushing a segment of older workers in the opposite direction. This trend affects workers in traditionally stable, well-compensated professional roles that previously offered longer career trajectories than physically demanding jobs.

Exposure levels mirror mid-career workers

Workers ages 55 and older face AI exposure levels comparable to their mid-career counterparts. Using the Digital Planet AI Exposure Index from Tufts University, researchers measured how well AI can perform tasks within specific occupations. The index incorporates assessments of large language model capabilities, machine learning suitability, and AI's ability to replicate human abilities required for various jobs.

Occupations with the highest exposure scores—above 70 on a 100-point scale—typically involve data work combined with coding. The lowest-scoring roles, below 30, require physical interaction with machinery or people.

Job exits rise in high-exposure roles

Before ChatGPT's launch, workers in AI-exposed jobs actually had lower exit rates than those in less-exposed positions, likely due to less physical demands and higher compensation. The research controlled for demographic factors including race, education, marital status, and earnings.

Since late 2022, however, that advantage has eroded. The study found significant increases in transitions to unemployment specifically—workers actively seeking new positions but unable to find them—rather than voluntary retirement or labor force exit.

For computer programmers, the analysis predicts a 25 percent increase in work exits compared to pre-ChatGPT levels. Painters, with minimal AI exposure, saw only a 2 percent increase. Despite these shifts, high-exposure jobs still maintain lower overall exit rates than low-exposure positions, though the gap has narrowed considerably.

Two paths to exit

The research identifies potential mechanisms driving these patterns. Workers may face involuntary displacement as AI automates core job functions. Alternatively, some may voluntarily exit rather than adopt unfamiliar technology late in their careers. Survey data indicates that while 18 percent of workers ages 50-64 use generative AI, adoption rates lag significantly behind younger cohorts. An AARP survey found 28 percent of workers 55+ view AI solely as a threat.

The findings draw on Current Population Survey data spanning January 2014 through April 2025, tracking individuals' employment status over one-year intervals. The analysis excluded pandemic-affected periods from March 2020 through December 2021.

Researchers at the Boston College Center for Retirement Research first reported these findings, cautioning that early AI impact assessments require careful interpretation as the technology landscape continues evolving rapidly.