Brookings says 15.6 million STARs, workers without a four-year degree, are in roles that sit in the top quartile of observed AI task exposure. That is one-fifth of the nation’s 70 million STARs, and 43% of all U.S. workers in the highest-exposure tier, Brookings reported in April 2026.

The exposed jobs are not scattered evenly across the economy. They cluster in administrative support, clerical work, and customer service, the exact sorts of positions that have traditionally given workers a first foothold in white-collar careers. Six Gateway occupations alone account for almost 8 million STARs in highly exposed work, Brookings reported in April 2026.

Location matters too. Brookings found especially high shares of AI-exposed STARs in Florida metro areas including Palm Bay (35.5%), Cape Coral (34.7%), Jacksonville (33%), North Port-Sarasota (32.7%), Orlando (32.2%), and Tampa (32.2%). Midwest metros looked less exposed, with Cincinnati (24.1%), Milwaukee (24%), and Des Moines (26%) lower on the list, Brookings reported in April 2026.

That geographic split is a reminder that this is not just an automation story. It is a labor-market story about where young workers start, and what happens when the starting point stops teaching the skills that lead somewhere better.

Entry-level work has always pulled double duty. It gets routine tasks done, and it trains people for more advanced roles later. Brookings argued in January 2026 that AI threatens both functions at once: if software absorbs the repetitive tasks, the training function can vanish with them.

The career ladder problem shows up in the pathways between Gateway jobs and higher-paying Destination jobs. Brookings said almost half of those pathways are highly exposed to AI, and only 51% are not highly exposed, Brookings reported in April 2026. A worker may keep the first job, then find the next step narrowed.

The risk is steeper for people with less room to absorb a hit. Brookings found 23 million STARs have low adaptive capacity, meaning they have limited ability to weather displacement and move into new work. Those workers make up 68% of the nation’s low-adaptive-capacity workforce, Brookings reported in April 2026.

Brookings also identified about 3.5 million STARs who are both highly exposed to AI and have low adaptive capacity. That group represents 67% of all U.S. workers in that doubly vulnerable position, Brookings reported in April 2026.

MIT Sloan adds an important counterpoint. As of December 2023, AI had not caused major changes in total employment, and losses in highly exposed roles were largely offset by gains elsewhere and by hiring growth at firms using AI to become more productive, MIT Sloan found in October 2025. That data predates the rapid rise of generative AI, so it is not the final word. It does, though, cut against the more apocalyptic takes.

The most practical message in the research is blunt: people who keep building skills are in a better position than people who wait. Brookings researcher Dimitris Papanikolaou said workers with the strongest incentive to invest in new skills seem to benefit most from AI-driven reallocation, and that skill-building also helps insulate them from displacement, Brookings reported in November 2025.

MIT Sloan’s advice runs in the same direction. The researchers said firms should encourage hands-on use of AI now, choose tools that fit the work, and push time toward tasks where people have the edge, such as critical thinking, novel problem-solving, and client relationships, MIT Sloan reported in October 2025. The point is not to worship the machine. It is to use it where it helps and keep humans on the work that still requires judgment.

Brookings goes further and suggests a structural answer: redesign early-career roles so learning is part of the job, not a lucky side effect. In January 2026, the think tank compared the idea to medical residencies, where learning is built into the role and taxpayers help subsidize the training through Medicare because society benefits from competent doctors.

That logic could be extended to AI. Brookings proposed an AI workforce reinvestment fund in which firms that automate away entry-level roles would contribute to pooled training programs across industries, and it pointed to the U.K.’s apprenticeship levy as a “use it or lose it” model for employers, Brookings reported in January 2026.

The evidence does not show mass AI layoffs already sweeping through entry-level work. Brookings said as much in January 2026, and MIT Sloan’s data also showed that total employment had not been broadly hit as of December 2023. That matters, because a panic story can be easier to sell than a messy one.

What the data does show is exposure, unevenly distributed. It also shows that AI can shrink some roles while expanding others, depending on how much of the job it can do. MIT Sloan found that when AI can perform most of the tasks in a role, the share of people in that role within a firm falls by about 14%; when AI affects only a few tasks, employment in that role can grow, MIT Sloan reported in October 2025.

That distinction is the one worth keeping in mind. AI is not only a replacement technology. It is a job-design technology too, and the result depends on whether employers use it to strip work down or to shift people toward tasks that still need a human being attached to them.

The researchers closest to the issue are careful about prediction for a reason. Papanikolaou said in November 2025 that five-year forecasts are “what ifs,” not settled answers, and that no one can say with confidence what AI will be able to do by then, Brookings reported. That uncertainty cuts both ways. It is a warning against overclaiming and a decent argument for preparing before the ladder loses another rung.

• Brookings found that 15.6 million STARs are in roles highly exposed to AI, Brookings reported in April 2026.
• Brookings said 23 million STARs have low adaptive capacity, Brookings reported in April 2026.
• MIT Sloan found that AI can be associated with lower employment in roles where it performs most tasks, and higher employment where it affects only a few, MIT Sloan reported in October 2025.