Tech Explained: AI Could Disrupt Millions of Jobs. Lawmakers Are Preparing in Simple Terms

Editor’s Note: This article appears in Governing’s Q1 2026 Magazine. You can subscribe here.

New Jersey Sen. Troy Singleton is worried about AI. More to the point, he’s concerned that the quickly evolving technology will put his working residents out of a job — some of them, perhaps permanently.

“We welcome innovation here,” Singleton tells Governing. “But it can’t be at the expense of our families and our community.”

Generative AI tools can already perform the core tasks of several existing jobs, including medical transcriptionists and customer service representatives. As a result, the Bureau of Labor Statistics anticipates that between 2023 and 2033, we will see a roughly 5 percent drop in employment in these careers. The bureau also sees a threat to various insurance roles; insurance companies may hire fewer claims adjusters, instead relying on drones to take photos of damage and AI to analyze the images and create initial payout estimates. The list of potentially disrupted jobs goes on, most of them white collar.

With the potential threat of AI-related job loss looming, Singleton is trying to prepare the state to better weather these changes — and better prepare workers for the in-demand jobs of the future. The senator’s new bill would require AI developers, as well as companies that lay off their workers and replace them with AI, to contribute to a fund for helping displaced workers retrain and find new employment. His bill would also lengthen unemployment benefits for those losing jobs to AI.

“While we can’t predict the future, it’s about building systems to help workers not only adapt quickly and safely, but have the dignity they deserve as the economy continues to change,” Singleton says. “Technology is going to change how we work, and that’s inevitable, but history demonstrates that prepared workers benefit most.”

New technologies have always brought about both new jobs and job losses. In that regard, experts say, AI is nothing new.

Industrial robotics, which came into use in the late 20th century, created a dip in manufacturing jobs. Service workers lost jobs at that time, too, likely because households were spending less money amid the shifting job market, per research from the American Economic Association.

That particular shift hit men harder than women, the opposite of what we’re likely to see with the coming AI turn. Men were far more likely to be in jobs focused on physical tasks that were at risk of robotic competition. Meanwhile, women were more likely to be in jobs that emphasized cognitive and administrative skills — the kind of jobs that may now face greater exposure to AI. Entry-level white-collar jobs have traditionally offered people a path to social mobility. If AI blocks the way for those workers, society could become more stratified, says Massachusetts Institute of Technology (MIT) researcher Isabella Loaiza. “If these [jobs] go away, people can’t move up … that’s one of the things that keeps me up at night,” she says.

In certain fields, generative AI tools are already being used to help perform entry-level work. Some tech companies report using generative AI tools to write code, with experienced software engineers then checking and fixing the tool’s output. That could be one factor behind a reported decline in employment for young, early-career software engineers. A Stanford study found hiring for such professionals dropped 20 percent from a hiring high in 2022 to 2025 (though Loaiza warns to be careful of attributing all recent declines in white-collar hiring entirely to AI, as opposed to broader economic conditions).

The Bureau of Labor Statistics anticipates that generative AI will create weaker hiring for certain legal jobs.

AI will be able to take over tasks like sorting through troves of research, summarizing findings and drafting legal documents. That’s likely to lead to slower job growth for paralegals and legal assistants, but not eliminate those roles. Meanwhile, the bureau expects the technology to increase productivity for human lawyers, who will remain necessary to understand ambiguities and context, as well as to provide the human relationship and trust important to some client services.

A Brookings and Centre for the Governance of AI report looks at which workers in highly AI-exposed jobs would be best able to weather job loss, should it occur. The report considers factors that have in the past correlated with workers’ ability to endure job displacements and find new work, and considers which people in highly AI-exposed jobs may have these adaptive advantages.

In general, the AI-exposed workers who will be most able to bounce back from job disruptions will be well-paid employees who have the savings to keep them going while they job hunt, as well as the diverse, transferrable skill sets and big professional networks to help them find new work, says Brookings Metro senior fellow Mark Muro, report co-author. Think lawyers, financial managers and software developers.

People also tend to have better prospects for finding new work if they’re young and live in areas densely populated enough to have many local job opportunities. The report finds that 71 percent of the U.S. workers with the most occupational AI exposure have above-average ability to adapt — that is, they “are among those best positioned to make a job transition if displacement occurs.”

But other kinds of workers will struggle to bounce back, if past trends are anything to go by. This group includes administrative assistants, medical secretaries, payroll clerks, and other clerical and administrative employees, per the report. They tend to have specialized skill sets that don’t translate easily to other jobs, and are in jobs that afford them fewer liquid savings. The latter makes it harder to weather a loss and requires people to settle for whatever job they can find quickly rather than hold out for a better position.

Another resiliency challenge comes for workers who live in less populated areas, which will have fewer job opportunities. The Brookings report found many of these high-risk workers live in small metro areas and college towns in the Mountain West and Midwest. Age is another factor. During the Great Recession, people aged 55-64 who lost jobs were significantly less likely to find new work than those aged 35-44. The report notes that “Older workers are less likely to retrain, relocate or switch occupations compared to their younger counterparts.”

Policymakers should focus their interventions on the AI-exposed workers who may need the most help recovering, Muro says. The report estimates 4.2 percent of workers in highly AI-exposed jobs — 6 million people — have low ability to adapt to job loss, and the majority of people in these roles are women.

“We saw what happened with industrial automation, with many people losing their jobs, receiving very little support and never really reconnecting with the labor market,” Muro says. “It would be a catastrophe to see that repeated now at what could be a much larger scale.”

Many experts believe we’re in a critical moment for policymakers to sway how the technology develops and how the impacts hit workers. Early action could avoid the mistakes of the past.

It’s also true, however, that we don’t yet know how AI will create new opportunities in the job market. MIT researcher David Autor estimates that 60 percent of U.S. jobs today did not exist in 1940, indicating just how much jobs change: “Most contemporary jobs require expertise that didn’t exist back then, and was not relevant at that time,” he says. (Of course, his research notes, factors like population shifts also contribute to the rise of new jobs — such as an aging population creating greater demand for health-care services.)

Still, job creation doesn’t always exactly offset job loss. Autor’s study finds that between 1980-2018, automation in the U.S. replaced more jobs than it created. Also, as technologies shake the job market, there’s no guarantee that the people who lost jobs are the same ones who will move into the newly created roles — one reason some policymakers are considering how to help smooth the transition for displaced workers.

Job markets can be very localized, making state and local governments well-positioned to foster programs that train people for careers in demand locally. Some of those in-demand careers we already know to be AI-resilient.

Software cannot take over the hands-on activities involved in blue-collar trades and patient-care work, at least unless robotics advance further or AI-integrated robots become pervasive. Several states are working to help their constituents gain the skills they need for these jobs. Apprenticeships are one major piece of that puzzle. “Apprenticeships and skilled trades are among the most resilient careers in an AI-enabled economy,” says a spokesperson for the California Labor and Workforce Development Agency. In the last fiscal year, the state increased annual building and construction apprenticeship grants from $3 million to $20 million.

States like Texas are making similar bets. In November 2025, the Lone Star State had more than 40,000 apprentices in different skilled trades. Last year, state lawmakers across the country reportedly introduced roughly 250 apprenticeship-related bills and 185 bills regarding career and technical education. A 2025 Washington law makes exploratory technical education courses available to kids as young as sixth grade. The courses include areas like manufacturing, welding, construction, culinary arts, automotive technology, early education and health-care diagnostics, as well as areas like coding and business management.

New Jersey Sen. Singleton’s bill also looks to apprenticeships to help people find good jobs. The bill calls for investing in “union-led apprenticeships” and “enhancing community college programs” to prepare residents for “jobs that are resilient, dignified and future-proof.” The legislation does not specify what those jobs would be, but Singleton expects strong demand in areas like health care, cybersecurity, advanced manufacturing, life sciences and clean energy.

Some education opportunities should likely be aimed at granting people the ability to use AI successfully, some experts say. In Tennessee, a public university created an “AI for All” minor for students on any degree path; the state’s AI Advisory Council highlighted this effort as a positive step toward preparing students for the future workforce. Texas, in turn, plans to incorporate lessons on generative AI into a program that helps adults without high school degrees, English fluency, or basic math or literacy skill up and find work.

Progressive think tank The Century Foundation recommends investing in education and training programs that emphasize the kinds of skills that are both AI-resistant and necessary for using technology while avoiding falling for its mistakes, such as ethical reasoning, critical thinking and field-specific expertise. Students learning about generative AI should not just be taught to craft effective prompts but also to evaluate the AI’s responses, the think tank recommends.

Singleton agrees. Many fields may use AI, but still need a human to operate that technological tool and provide the skills AI cannot replace, he says. “We have to invest early in digital and, frankly, human skills,” Singleton says. “[AI] cannot easily replace problem-solving, teamwork, adaptability and leadership; those things are still uniquely human.”

AI can mimic conversational language, but it still cannot truly think. Ethical and moral judgements, critical thinking, leadership, teamwork and fostering trust remain the domain of humans, some researchers and policymakers believe. That may make jobs like judges and client-facing financial roles, where customer trust is essential, safer bets.

Whether AI will encroach on other traditionally human skills is more disputed. Only humans can be genuinely empathetic and emotionally intelligent, but AI can mimic empathetic-sounding language; the impact of AI on jobs that require empathy will depend on how useful employers or consumers deem that quality to be.

Similarly, while AI cannot do original thinking or imbue artistic work with genuine meaning, AI models trained on people’s creative works are already able to produce illustrations and stories. The Hollywood writers’ strike succeeded in getting major movie studios to agree that generative AI would not be used to replace writers. Some strikers had feared studios would lay off most writers and relegate the remaining ones to cleaning up algorithmically generated scripts, threatening both the quality of the product and writers’ job stability and compensation.

Part of the question of job replacement is not whether AI will necessarily do a task better than humans, but whether it will do it sufficiently well and for sufficiently less cost that employers decide the trade-off is worth it. But employers — and policymakers — do have a choice in how this turns out, experts say.

“A lot of this risk [to jobs] will be realized or not depending on the decisions we make to implement the technology,” Loaiza says. “The path is not predetermined, and it’s up to us to decide what we want to do with this technology.”