New research from Boston College’s Center for Retirement Research finds that workers aged 55 and older in AI-exposed occupations are leaving their jobs at higher rates since the arrival of generative AI, a shift that collides directly with a Social Security trust fund projected to run dry by late 2032.

The workers most vulnerable to AI displacement tend to be college-educated, higher-earning professionals, the same cohort that may face the steepest benefit cuts when Social Security reform finally arrives. For investors focused on capital preservation, the convergence of forced early retirement, shrinking entitlement promises, and inflation-eroded savings makes the retirement math considerably harder than Washington acknowledges.

The paper, authored by economics professor Geoffrey Sanzenbacher, draws on Current Population Survey data and AI-exposure scores developed by Tufts University’s Digital Planet initiative. Its central finding is simple: before the launch of OpenAI’s ChatGPT, older workers in AI-exposed roles were significantly less likely to leave their jobs. After the launch, that pattern reversed. Those same workers became somewhat more likely to transition out of work, including into unemployment.

CNBC reported that Sanzenbacher described the effect as statistically significant and, for some occupations, “quite large.” The transitions are split roughly evenly between involuntary job loss and voluntary departures, a distinction that matters, because it suggests the pressure is not just layoffs. Some workers appear to be choosing to leave rather than adapt to AI-saturated workflows.

Who Gets Displaced, and Why It Matters for Retirement Policy

The demographic profile of the affected workers is not what casual observers might expect. The research found that older workers most susceptible to AI-driven career disruption tend to be white, are much more likely to hold a college degree, and earn more than their peers in low-AI-exposure roles. These are not assembly-line workers or retail clerks. They are knowledge workers, professionals, and managers, the people whose longer careers and higher earnings have historically subsidized the rest of the Social Security system.

That creates a policy problem. As Sanzenbacher wrote in the paper, “AI exposure may reduce the gap in career length between low- and high-paying jobs.” In plain language: the well-paid professionals who used to work the longest may now be leaving earlier, which shrinks the payroll-tax base and accelerates Social Security’s funding gap.

Related Posts
Ultra-rich investors cut stocks and build cash reserves

The timing is brutal. The program’s trustees project the trust fund that helps pay retirement benefits may run out in late 2032. The last major reform came in 1983, when lawmakers gradually raised the retirement age from 65 to 67. Raising it again is one option on the table. But if AI is already shortening careers for higher-income workers, pushing the retirement age higher could leave a growing number of displaced professionals stranded between their last paycheck and their first benefit check.

“There’s a high probability that higher-income people see a bigger benefit cut than lower-income people from whatever happens with Social Security next. These are the very people who therefore need to work longer.”

That warning from Sanzenbacher frames the core tension. The workers who most need extended careers to offset likely benefit cuts are the same workers AI may be pushing out early. The research explicitly urged policymakers considering retirement-age changes to account for AI’s effects on career length.

The Perception Gap Among Older Workers

A separate AARP survey of 1,015 adults aged 50 and over found a divided outlook. Just 24% said they see AI as a threat to their line of work. Nineteen percent called it an opportunity. The largest group, 37%, said it was both. That ambivalence tracks with a labor force that senses disruption but has not yet fully absorbed its implications.

As we explored in our coverage of why millions of older Americans can’t afford to stop working, the financial pressure on late-career workers has been building for years, well before generative AI entered the picture. What the Boston College research adds is a specific, measurable accelerant.

Joint research from AARP and LinkedIn offered a partial counterweight: 49.4% of older workers occupy roles insulated from generative AI disruption, compared with 42.2% of younger workers. The reason is that older workers’ jobs are more likely to require skills AI cannot easily replicate, collaboration, judgment, and leadership. But “insulated” is not the same as “immune,” and the research on job transitions suggests the insulation is already fraying at the edges.

The AI Adoption Reality

A Monster WorkWatch report surveying 1,504 workers found that 42% don’t use AI at all. Among those who do, the most popular applications are basic tasks: email, scheduling, and writing support. Others use it for coding, automation, data analysis, job applications, or creative work. The gap between AI’s theoretical reach and its actual workplace penetration remains wide.

Vicki Salemi, a career expert at Monster, framed the adaptation challenge in practical terms:

“When you can show you possess strong soft skills coupled with the ability to evolve and grow with new technology, it can be a green light for your candidacy.”

That advice is sensible. It is also insufficient for a 60-year-old financial analyst or marketing director whose entire department just adopted tools that can do 70% of the job in a fraction of the time. The question is not whether older workers can learn new tools. It is whether the economics of retraining pencil out when the career runway is short and the employer’s incentive is to hire someone younger who already speaks the language.

The geographic dimension adds another layer. New York Post reporting from 2019 on senior labor-force participation showed that older workers in manufacturing-heavy Rust Belt regions and parts of the South already face grim job prospects after automation and globalization hollowed out local industries. Gary Burtless, a senior fellow at the Brookings Institution, described the problem bluntly: “A lot of them are one-industry towns. And if that industry has been hit hard, that’s going to be a problem for younger workers and older workers.” AI adds a white-collar version of the same dynamic.

What This Means for Retirement Security, and for Gold

The investment implications run deeper than the labor-market headline suggests. Consider the chain of effects:

  • Higher-earning workers exit the labor force earlier, reducing lifetime savings and payroll-tax contributions.
  • Social Security’s funding gap widens, increasing the probability of benefit cuts, means-testing, or retirement-age increases.
  • Displaced professionals draw down retirement accounts sooner, potentially selling financial assets into weaker demand.
  • The political pressure to “do something” about retirement security grows, raising the odds of fiscal interventions funded by borrowing or money creation.

For readers tracking the Gen X retirement crisis and inflation’s toll on savings, the AI displacement research adds a structural headwind that most retirement-planning models do not incorporate. The standard advice, work longer, save more, delay claiming benefits, assumes the option to work longer actually exists. If AI is compressing career spans for the professional class, that assumption weakens.

The fiscal math matters for metals investors specifically. Every pathway to closing Social Security’s funding gap involves either higher taxes on a smaller base of high earners, benefit cuts that reduce consumer spending, or deficit spending that adds to an already unsustainable debt trajectory. None of those outcomes are friendly to the purchasing power of the dollar. All of them reinforce the case for hard assets as a hedge against policy-driven erosion of real wealth.

This is not a distant hypothetical. The 2032 trust-fund exhaustion date is less than seven years away. And as tech CEOs have openly acknowledged, AI-driven workforce reductions are not slowing down. The intersection of accelerating displacement and decelerating entitlement funding creates a structural problem that no amount of reskilling advice can fully solve.

The Deeper Signal

Sanzenbacher’s research identifies a mechanism that deserves more attention than it is getting. AI is not just a productivity tool or a labor-market disruptor. It is a force that may reshape the actuarial foundations of the retirement system itself. If higher-income workers lose years of career length, the entire architecture of Social Security, built on the assumption that those workers contribute the most, for the longest, comes under strain from a direction policymakers have barely begun to model.

For capital-preservation-minded investors, the takeaway is not panic. It is recognition that the retirement system’s assumptions are degrading in real time, and that the political responses, when they finally come, will almost certainly involve some combination of benefit haircuts and fiscal expansion. Both paths favor tangible stores of value over promises denominated in a currency whose custodians face impossible arithmetic.

The workers being displaced by AI are learning a hard lesson about the difference between income and wealth. For retirees already navigating inflation’s drag on their savings, that distinction was never abstract. It is becoming less abstract for everyone else, too.

When the system’s promises depend on assumptions that are actively breaking down, the assets that do not require a counterparty tend to hold up best.