Last recruiting cycle, Kevin Li, a rising senior at Duke University, applied to somewhere between 150 and 200 internships. But by his classmates’ standards, he told me, his effort was modest. The students who actually land spots are “probably sending out hundreds of applications,” spending hours at networking events, or “trying to DM dozens, even hundreds of people” on LinkedIn and other platforms. And “you have to start early,” even before applications themselves open. Li got two interviews and zero offers. He spent the summer doing research at Duke instead.
The AI-induced white-collar employment apocalypse isn’t quite here. Last March, Anthropic identified business, finance, computing, and law as among the fields where large language models could theoretically displace the most work. But since ChatGPT became available to the public in 2022, the ranks of financial analysts, software developers, and paralegals have each grown 8 to 9 percent.
Still, the market for interns is another matter. Internship postings on Handshake, the hiring platform, fell at least 10 percent for each of the last three recruiting seasons. The platform says it saw 30 percent of tech internships and 42 percent of professional services internships erased between 2023 and 2025. And the search is getting much more competitive, too. Internships posted on Handshake in the 2025 academic year drew an average of 109 applicants, up from 62 the year before. The internship apocalypse is here.
In many white-collar settings, very little of what an intern does is too important or complex for the latest models to handle. Congressional interns, for instance, have spent decades sorting mail from constituents, answering phones, and giving tours of the Capitol. AI excels at the first two; the third is essentially make-work. Perhaps no kind of worker, then, is more vulnerable to displacement.
Internships are supposed to give college students a few months of work experience to help them land a full-time job after graduation. And for employers, they’re too important to do away with entirely: They offer a low-stakes chance to evaluate potential junior hires before committing to them. Caught between these two forces, what becomes of the intern?
This summer, many members of Gen Z feel their internships—never the most productive of arrangements—have become a farce, drained of anything resembling meaningful work. Many of those lucky enough to land spots are encouraged by their bosses to delegate nearly everything to AI. Jack Vu, a rising sophomore at Rice University, is spending the summer at a private equity firm in Houston. His boss, he says, “is as into AI as anybody on the planet”—he “thinks that every second you’re not using AI is a waste.” By his own estimate, he delegates 90 percent of his work to Claude, Anthropic’s AI assistant. The outputs come back “mostly ready to go.” For the remaining hours of an eight- or nine-hour day, he scrolls X, reads books, and, for the duration of the World Cup, watched soccer. “Anything you can think of to kill time,” he told me.
A good intern finds more things to do, keeping an eye out for problems no one asked them to solve. But even Vu’s self-assigned work mostly consists of waiting for the models to return their outputs. “Claude isn’t really separable from the work,” he told me. “If I’m doing research on some thesis or trying to think of some model to target a certain thing, like, I’m using Claude the whole way through.”
Other bosses are simply unaware of what the technology can do. In his first week on the job, one intern—a coder at the quantitative trading firm Susquehanna International Group—told me his boss had assigned him a dozen tasks, scattered across a 15-year-old codebase he’d never seen. (He requested anonymity, as he was not authorized to speak with the press.) But he soon discovered that the tasks he’d been hired to do could be handled start to finish by Claude Code, Anthropic’s coding tool.
In two weeks, he cleared what his supervisor had budgeted as several months of work. That astonished his boss. So, the intern did the only sensible thing and slowed down, rationing the backlog to keep up appearances, working maybe half a day and killing time for the other half.
Once, he aimed Claude at a performance bug that caused his firm’s trading software to lag. Claude returned a clarifying question: Which of the possible fixes did he want applied? Then the model fixed it. At the next team meeting, a senior engineer said the now-patched bug had been bothering him for six years. “No one else on my team, because they’re all, like, adults, have really thought to do this,” he told me. They’re akin to “your boomer grandma” who “thinks you’re so smart for, like, Googling a basic question.”
What’s left for him to do with the rest of the day? He’s building a bot that lets colleagues check whether a trade has gone out to the clearing firm without emailing his team and waiting. But he’s hit a wall. “I could be way more useful, but I’d need to learn how to use some tool or another tool,” he told me. “And in order for me to learn that, someone has to teach me because it’s human knowledge and not, like, something I can look up.” Before AI, a few weeks of onboarding would’ve been enough to make an intern useful through the summer. Now, to use him at his actual capacity, “they would need to invest more time in me.” And the people with the relevant knowledge are busy with other tasks.
So, he’s become a reader of the Western canon on the clock, he told me. He is working his way through a new translation of The Epic of Gilgamesh. When Gilgamesh palls, he switches to Alexis de Tocqueville’s Democracy in America. He also studies a more contemporary canon: Instagram Reels.
Plenty of interns aren’t outsourcing all their work to AI: Finance and tech are two fields where AI adoption was high to begin with. Personally, I spent the summer as an intern at this very magazine. Mercifully, my experience was nothing like what’s described here.
But advances in AI have already disrupted many internships, and the disruption ahead may be greater still. In the process, as interns have less meaningful work than ever, employers will be forced to ask what the whole exercise is actually for. The intern’s purpose has always been “fuzzy,” David Deming, an economist at Harvard’s Kennedy School and co-director of its Project on Workforce, told me. Employers might “actually get some valuable work out of them”—but just as importantly, they “get a little bit of a look at them,” and can later decide whether to bring them on full-time. “My internship exists for the purpose of recruiting future employees,” the Susquehanna intern told me. Nearly everyone he’s met at the firm is a former intern; his own boss was one.
If internships are primarily a recruiting tool, employers have little reason to abandon them, no matter how little work interns end up doing. And with the grunt work automated, Deming said, some interns might actually receive more meaningful, open-ended assignments. “Maybe instead you give the internships a kind of passion project, and you tell them: Find some problem you think should be solved in the organization, and you go off and try to solve it.”
Some employers are taking Deming’s advice. Janvi Shah, co-founder and CEO of the ecommerce startup Hue, told me that her data intern once assembled reports for over a hundred clients by hand. Now the intern automates that work, spending the recovered hours on “higher-level things” like “looking for trends.” But interns in this category need not offer their firms anything of immediate value. As Shah told me, “A huge portion of our thought process with interns is the talent identification.” And if AI could do everything the interns do, she thinks she’d keep them on anyway.
Even as the number of internships is shrinking, the share of interns who get hired full-time is rising. According to the National Association of Colleges and Employers, companies converted 63 percent of their 2024–25 interns into full-time hires, the highest rate in five years. That combination suggests employers aren’t cutting internships across the board. Instead, they may be cutting the ones that exist to extract cheap labor, while retaining the ones that fill the recruiting pipeline.
Meanwhile, the marginal internship, the kind that existed only because the work needed to be done, may not last. Abhishek Agrawal is chief of staff at Innovaccer, a San Francisco-based healthcare-data company that hires 80 to 100 engineering interns a year. The company is growing fast—its annual recurring revenue (ARR) has roughly quadrupled over four years, Agrawal told me—but its appetite for interns hasn’t grown with it. “You don’t need four people to do the same job that was done earlier,” he said. “Now one person can do that job.”
As long as there’s demand for human labor in the AI age, there will be demand for interns. But fewer internships will be available, and those who secure them will have far better odds of turning them into permanent work. Internships will be rarer, and as employers demand more experience for entry-level jobs, worth much more.
At least in the near-term, the path into America’s white-collar workforce will grow ever harder to navigate. Research has found that a summer of paid work outside one’s field has virtually no sway over a recent graduate’s appeal to employers. And, in business fields, a relevant summer internship counts for more than even your choice of major. But as candidates with internships on their resumes become scarcer, recruiters may come to lean more heavily on a candidate’s college major, broader history of student employment, or, God forbid, GPA. (Grade inflation, of course, complicates the last of these.)
The internship as a mass rite of passage into the labor force is likely ending. It’s cold comfort for the students struggling through the transition, but maybe some disruption was overdue. It has never been obvious why a 10-week stint fetching coffee and reformatting PowerPoint decks should determine who gets hired at 22. Couldn’t we undergraduates, already in front of screens so much of the year, be doing something more meaningful with our summers?
