You have worked in the same office for close to twenty years. Then one quarter your manager starts using the word “efficiency” a lot, and by the following spring your job does not exist anymore. Nobody says the words artificial intelligence out loud. They rarely have to.
A 2026 analysis from the Center for Retirement Research at Boston College found that since ChatGPT launched in November 2022, workers 55 and older in the most AI exposed occupations have become significantly more likely to leave those jobs. Some of that is early retirement. A lot of it is not. The researchers found the increase was driven largely by transitions into unemployment, not by workers choosing to stop working.
That timing could not be worse. More than half of workers over 50 are laid off or pushed out of career jobs before they choose to leave on their own, and once that happens, AI powered hiring tools have been shown to disadvantage applicants over 45 during the first round of resume screening. Lose the job, then lose ground finding the next one, for reasons that have nothing to do with skill.
Here are 15 jobs where that squeeze is showing up first, and why.
Computer programmers
Programmer jobs are projected to decline 6 percent through 2034, as companies use AI to automate routine coding tasks and shift the more complex work to software developers instead. That is a shrinking pie for a role that used to be a stable, long career.
The Boston College research on older workers singled programmers out directly. Once ChatGPT became widely available, the rate at which older programmers left their jobs jumped by more than 25 percent compared with the years before, one of the largest increases of any occupation studied. Painters, by comparison, saw almost no change at all. The gap says a lot about which kinds of work AI is actually reaching first, and it is not the physical jobs.
A programmer who has spent two decades writing code for one industry now has to compete for far fewer openings, often against younger candidates willing to work for less and companies eager to test whether an AI coding assistant can do the job for a fraction of the cost.
Accountants and auditors
On paper, this looks like a growing field. The Bureau of Labor Statistics expects 5 percent employment growth for accountants and auditors through 2034, faster than average, driven by a growing economy and increasingly complex tax rules.
But growth in the field overall does not mean older workers inside it are safe. The same Boston College research found that exits from the workforce among older accountants and auditors climbed by more than 22 percent after ChatGPT's release, a jump nearly as steep as the one seen among programmers. AI tools are increasingly handling reconciliations, first pass audits, and routine reporting, the kind of work senior accountants used to hand down to junior staff. That leaves fewer places for an experienced accountant to land once their role is restructured, even while the overall job count keeps climbing.
Firms are still hiring accountants. They are just hiring fewer of them at the senior end of the pay scale, and reorganizing teams around a smaller group of people who can oversee what the software produces rather than a larger group who once did that work by hand.
Executive secretaries and administrative assistants
Secretaries and administrative assistants are expected to see little or no employment growth through 2034. The Bureau of Labor Statistics points directly at the cause: automated phone systems and AI scheduling assistants are steadily reducing how many people a company needs in this role.
Within that broader group, the specific job of nonmedical secretary and administrative assistant is projected to shrink by about 31,000 positions over the decade, even as medical secretaries, a separate and growing specialty, pull the overall numbers up.
This is a role that has historically rewarded tenure, institutional memory, and the ability to manage a demanding calendar without being asked twice. Those are exactly the skills that do not show up in a job posting anymore, because an AI assistant now handles the scheduling and a chatbot answers the routine questions.
Executives who once relied on a longtime assistant to anticipate their needs are increasingly told the AI tool can do the same thing for a fraction of the cost, even if it cannot actually read the room the way a person who has worked alongside them for a decade can.
Customer service representatives
Nearly 154,000 customer service jobs are on track to disappear by 2034, a 5.5 percent decline, even though hundreds of thousands of openings will still appear each year as people retire or move on.
Self-service apps, chatbots, and AI phone systems now resolve a huge share of the simple questions that used to require a live person, which means fewer representatives are needed to handle what is left. The jobs that remain increasingly go to the workers who can manage the hardest, most emotionally charged calls, the ones a bot still cannot defuse.
That is not necessarily an advantage for someone who has spent twenty years handling routine customer service calls. Experience answering routine questions patiently does not automatically translate into handling escalated complaints, and companies restructuring their call centers are not always willing to retrain a longtime employee for the harder version of the same job. Call center veterans who built their whole career on staying calm and following a script now find that the script itself has been automated, leaving only the calls nobody, human or bot, wants to take.
Bookkeeping, accounting, and auditing clerks
This is one of the clearest cases of software directly replacing people. Employment for bookkeeping, accounting, and auditing clerks is projected to decline 6 percent through 2034, a loss of more than 94,000 jobs, as accounting software automates the reconciliation and data entry work these clerks used to do by hand.
Unlike accountants, who need a degree and can pivot into advisory or analysis work, bookkeeping clerks typically do not have an obvious next rung to climb to inside the same company. The tasks that made the job worth having, matching invoices, catching errors, keeping the books balanced, are precisely the tasks a modern accounting platform now does automatically and instantly.
For someone who built a career on being meticulous with numbers, watching that meticulousness get automated out of the job description is its own kind of loss, on top of the practical one. The openings that remain increasingly ask for comfort managing several software platforms at once rather than the manual bookkeeping skills a longtime clerk spent years perfecting.
Data entry keyers
No office occupation on the Bureau of Labor Statistics list of the steepest declines is shrinking faster than this one. Employment for data entry keyers is projected to fall almost 26 percent through 2034, a loss of nearly 37,000 jobs from a workforce of only about 142,000 to begin with.
Optical character recognition and AI powered document processing can now read a scanned form and populate a database faster and more accurately than a person typing it in by hand. There is very little room left for the argument that a human keyer adds something the software cannot.
Data entry has long been an entry point for workers without a college degree, including many older workers who came to office work later in their careers or after a layoff in another field. As the role disappears, that entry point disappears with it, closing off one of the more accessible paths back into steady office work.
Payroll and timekeeping clerks
About 27,000 payroll and timekeeping clerk positions, nearly 17 percent of the field, are projected to vanish by 2034, as payroll software increasingly handles calculations, tax withholding, and time tracking without much human review.
This job used to demand a sharp eye for discrepancies, someone who could catch a miscalculated overtime hour or a benefits deduction that did not look right. Modern payroll platforms flag those discrepancies automatically and route exceptions for approval rather than requiring a clerk to check every entry by hand.
Workers who have spent decades in payroll often built their value around exactly that kind of vigilance. When the software does the checking, the remaining openings tend to go to a much smaller team overseeing the system rather than a full staff running the numbers manually, and there simply are not enough of those oversight roles to absorb everyone the automation displaces. A payroll clerk who spent thirty years learning every exception in a company's pay structure now competes for one of a handful of remaining spots against people half her age who grew up managing software, not spreadsheets.
Bank tellers
Bank teller employment is projected to fall nearly 13 percent through 2034, a loss of about 45,000 jobs, continuing a decline that started with ATMs decades ago and has accelerated as banks add AI powered chat and voice tools for routine transactions.
Mobile deposit, automated fraud alerts, and AI chat assistants now handle a large share of what used to require a trip to the branch and a conversation with a teller. Branches are consolidating as a result, and the tellers who remain are increasingly expected to sell financial products rather than simply process transactions.
That shift favors newer hires trained from day one to pitch credit cards and loans, not necessarily the teller who has spent fifteen years building relationships with regular customers. Loyalty and familiarity, once real assets in this job, do not weigh much against a bank's push to do more with a smaller branch staff, and branch consolidation often means the surviving positions go to whichever employees are willing to relocate or shift into sales.
Claims adjusters, examiners, and investigators
The Bureau of Labor Statistics is unusually direct about the cause of this one. Insurers now use AI tools to assess photographs of property damage and generate payout estimates automatically, and the agency says that efficiency gain is projected to shrink employment for claims adjusters, examiners, and investigators by 5.1 percent through 2034, a loss of more than 18,000 jobs.
Judgment calls that used to require a trained adjuster physically inspecting a wrecked car or a flooded basement can now be handled, at least in the first pass, by a model trained on thousands of prior claims. The adjusters who remain are increasingly handling the complex or disputed cases the software flags rather than the routine ones.
An experienced adjuster's real skill, spotting the details that do not add up in a claim, is still valuable. The problem is there is far less routine work left to pair with it, and fewer entry points for that expertise to keep earning a full time salary.
Bill and account collectors
Bill and account collector jobs are projected to shrink by roughly 10.5 percent through 2034. Automated payment reminders, AI powered chat negotiations, and self-service payment portals now handle much of the routine outreach that collectors used to do by phone.
Collections work has always required a particular kind of resilience, the ability to have the same difficult conversation dozens of times a day without losing patience. That skill does not disappear just because a company decides a text message or an automated call can handle the first three attempts to collect a debt before a human ever gets involved.
What is left for human collectors tends to be the accounts that have already resisted automated contact, the hardest cases, with the least predictable outcomes. Fewer of those cases exist per collector than the volume of routine collections work that used to keep a full team employed, which means smaller teams handling a tougher caseload rather than the same number of people doing easier work.
Interviewers
The workers who conduct surveys, screen applicants, and gather information for research and government programs are facing a nearly 12 percent decline in employment through 2034. AI powered chat and voice tools can now conduct a structured interview and log the responses without a person on the other end of the line.
Government agencies and research firms increasingly deploy automated phone and web surveys for the kind of standardized questioning that used to require a trained interviewer working from a script. The tasks are repetitive enough, and specific enough, that generative AI systems can handle them with minimal oversight.
Workers who built careers on the rapport and consistency needed to get honest answers from strangers over the phone are finding fewer roles that value those specific skills, since a growing share of the interviews themselves are simply not being conducted by a human anymore. The remaining positions increasingly involve designing the survey instrument or reviewing flagged responses rather than doing the interviewing directly.
Paralegals and legal assistants
The paralegal field is projected to show little or no growth through 2034, a sharp change from the steady expansion it saw for decades. The Bureau of Labor Statistics attributes the flattened outlook directly to AI tools that can now handle legal research and document review far faster than a person can.
Reviewing thousands of pages of discovery documents for relevant facts used to be exactly the kind of labor intensive work that justified hiring a paralegal. AI systems can now scan that same volume of material in minutes and flag the passages that matter, cutting the number of research hours a law firm needs to bill or absorb.
Paralegals with deep experience in a specific area of law still bring judgment that software cannot replace. But with far less routine document review left to divide among a legal team, firms increasingly need fewer paralegals to support the same number of attorneys, and long tenure alone is no longer enough to guarantee a seat at that smaller table.
Editors
Editor jobs are projected to grow just 1 percent through 2034, essentially flat, with the Bureau of Labor Statistics noting that editors at traditional print newspapers and magazines specifically face declining employment as those publications lose ground to other formats.
Researchers studying AI's effect on older workers used an editor as their central example of a job with real automation risk. Correcting spelling and grammar, verifying facts, and evaluating whether an article reads clearly are all tasks generative AI can now do quickly, and the researchers noted the outcome for any individual editor depends heavily on whether that person adopts the new tools early or gets replaced by them later.
An editor who spent a career building judgment about what makes a story worth running is not easily replaced. But a newsroom under financial pressure does not always distinguish between the mechanical parts of the job and the parts that require real editorial instinct when deciding how many editors it can still afford.
Retail supervisors
More than 72,000 retail supervisor positions, first line managers who run the sales floor, are projected to disappear by 2034, a 5 percent decline, as retailers lean on AI powered inventory forecasting and scheduling software to do work that used to require a floor supervisor's judgment.
Predicting how much staff a store needs on a given day, flagging which products are running low, and building next week's schedule are all tasks a supervisor used to handle through experience and instinct. AI scheduling and forecasting tools now do a version of that work automatically, based on sales data rather than a manager's feel for the store.
A supervisor who spent years learning the rhythm of a particular store, the seasonal spikes, the regular customers, the staff who need more oversight than others, built expertise that does not transfer easily to a job posting. Retailers cutting management layers to lean on software tend to keep the newest, cheapest supervisors rather than the most experienced ones.
General office clerks
Employment for general office clerks, one of the largest office occupations in the country, is projected to fall 6.7 percent through 2034, a loss of nearly 178,000 jobs, the biggest numeric decline of any office and administrative role tracked by the Bureau of Labor Statistics.
This is a catch all role, filing, photocopying, answering phones, routing mail, handling whatever paperwork needs a human hand. AI powered document management and workflow software increasingly handles the routing and filing automatically, and shared inboxes with automated sorting cut into the need for someone to manually direct incoming requests.
Because the job covers so many small tasks rather than one specialized skill, it has long been a flexible landing spot for workers moving between careers or reentering the workforce later in life. As AI absorbs those small tasks one by one, that flexible landing spot keeps getting smaller, right as more experienced workers need somewhere to land.
The real gap shows up in reemployment, not just job counts
None of these jobs are disappearing overnight, and every one of them will still post real openings each year as people retire or move on. The harder problem is what happens next for the worker who loses one of these roles at 55 or 60. Among workers over 50 who expect their next job search to be difficult, age discrimination is the reason cited most often, ahead of any gap in skills or experience. That is the part a shrinking job count alone will never show.
