News

How AI Is Reshaping Career Risk for Workers Over 45

New research from the Center for Retirement Research at Boston College finds that older workers in AI-exposed occupations are now more likely to leave the workforce than they were before ChatGPT's launch in late 2022 — and much of that increase is driven by transitions into unemployment, not retirement. The findings were highlighted by CNBC on July 13, 2026.

Researcher Geoffrey T. Sanzenbacher analyzed data from the Current Population Survey alongside Tufts University's Digital Planet AI Exposure Index. The study found that computer programmers — one of the highest-exposure groups — saw work-exit rates rise by more than 25 percent between 2014 and 2025, while workers in a low-exposure role like painting saw an increase of just 2 percent. Accountants and auditors were also flagged among the high-exposure occupations. Before ChatGPT's arrival, workers in high-exposure roles had actually stayed employed longer than those in physically demanding, low-exposure jobs. That durability advantage has since weakened.

The exposure risk is unevenly distributed. Older workers who are most susceptible to AI-driven disruption tend to hold white-collar, higher-earning roles that require college degrees — the very positions that AI tools like code generators and automated auditing software now replicate most effectively. Workers in hands-on roles requiring direct physical labor or face-to-face interaction show far smaller increases in exit rates.

A separate concern emerges at the hiring stage. Research by Alysia Blackham at the University of Melbourne found that AI recruiting tools can embed age bias from the start of a job search. When prompted to recommend candidates for tech roles emphasizing energy and new ideas, ChatGPT suggested early-career professionals aged 21 to 30 and mid-career professionals aged 30 to 45 — with those over 45 left out entirely. Blackham concluded that AI in hiring is not overcoming existing bias but accelerating it.

Training access compounds the problem. An AARP Public Policy Institute survey of workers aged 50 and older found that 49 percent expressed interest in learning AI for work — but only 12 percent had actually taken any AI training or classes. That is a 37-percentage-point gap between intent and action, driven in part by employer inaction: 63 percent of respondents said their employers do not train workers adequately for AI, and 66 percent said their employers do not encourage AI training regardless of age.

What this means for job seekers

The picture is more nuanced than a simple threat narrative, and that matters for how experienced workers respond. Research from the Oliver Wyman Forum found that more than 40 percent of CEOs plan to reduce junior-level roles within one to two years as AI handles entry-level tasks — while keeping senior and mid-level positions stable. The judgment, institutional knowledge, and cross-functional experience that older workers carry are precisely what AI tools cannot replicate.

The practical implication: workers over 45 who close the training gap early stand to gain more than most from AI's reshaping of the labor market. If AI automates the junior tiers of a field, the roles left standing are those requiring accumulated expertise. Getting comfortable with AI tools — even at a basic level — reframes that experience as a force multiplier rather than a liability. Our review of the research suggests the workers most at risk are not those who are AI-exposed by occupation, but those who are AI-exposed and not adapting. If reskilling is on your radar, our guide to AI-proof career skills covers the capabilities employers are prioritizing right now.