Today, recent college graduates have a higher unemployment rate than the entire country’s workforce. The New York Fed puts the number at about 5.6% for graduates between 22 and 27 years old, as of the second quarter of 2026. The overall rate was only about 4.2%. Not once, in the 36 years the Fed has been tracking, has that number gone this direction before. Experienced workers are slowly fading away.
Ask almost any hiring manager what’s hardest about their job right now. You’ll hear some version of the same complaint: “nobody has the experience we need.” Three to five years direct experience, ideally. Someone who can ‘hit the ground running’. Here’s the problem with that wish. A job market can’t just make workers with three to five years of experience. Especially if it keeps cutting out the jobs where people get their first years of experience.
A Cycle Begins
- Each one of these facts shape this new system. Internship postings are down.
- Entry-level job listings ask for demand experience that entry-level workers just don’t have yet.
- Managers say they don’t have time to train anyone.
- Companies are promoting from within less than they used to and hiring more from outside.
When you read about any one of these things on its own, they look like unrelated HR trends. When you put them next to each other… they point to an economy that’s ‘defunding its own future supply of experienced people.’ Then we’ll end up wondering, a few years out, where all the experienced workers went.
The Cycle Itself
At Diversity Employment we took six separate sources side by side for this piece and didn’t find anyone else connecting them. So, here’s that system, one piece at a time.
The Vanishing Front Door
We start where people actually used to begin: The internship. Handshake’s own data shows that internship postings fell more than 15% between January 2023 and 2025. In tech and professional services, specifically, postings dropped by over 30% in that same stretch of time. Demand for internships went in the opposite direction. Applications per posting nearly tripled. They’re up from 43 in the 2022-23 cycle to 109 in 2024-25. In tech jobs, that number actually hits around 273 applications; for every single posted internship.
A Full Misalignment
Then it gets interesting. Asking employers directly, through NACE’s own surveys, most of them say they’re either maintaining or increasing their internship programs. NACE’s 2025 report put that figure above 70%, and its 2026 report projects a 3.9% increase in interns for this coming cycle. Employers aren’t lying when they say it… But what they say about internships and what they’re actually posting have split pretty far apart. Handshake’s data and NACE’s survey data are describing two different timelines. Only one of them can be true when a student actually goes looking for one.
The students who do land something are competing harder for it than any other recent class before them. By January 2025, 41% of the Class of 2025 had applied to at least one internship through Handshake. That number is up from 34% of the Class of 2023 at the same point in their college careers. ‘More people are chasing less job openings,’ is the plainest possible definition of a bottleneck, and it’s at the very front of the pipeline. It’s choking jobseekers before anyone even has a chance to build the experience employers say they want.
Entry-Level Jobs Got “Senior-ized”
Even the job postings that are still out there have changed their shape. Burning Glass Institute’s July 2025 report, “No Country for Young Grads,” documents a targeted decline in demand for workers with under three years of experience. It’s focused almost entirely at jobs where generative AI can do a big chunk of entry-level work. Marketing specialist, project manager, and financial analyst used to be classic first jobs. They were the kind with: a clear on-ramp built on research, first drafts, and basic analysis. The report calls this the “Expertise Upheaval.” AI is best at exactly the foundational tasks juniors used to learn by doing. Which means the traditional staffing pyramid, wide at the base and narrow at the top, is starting to flip.
The Numbers Don’t Lie
That shift is evident in unemployment data too. Unemployment for 20-to-24-year-olds with a bachelor’s degree rose from 5.2% in 2018-19 to 6.2% in the two years through mid-2025. Layoff rates for young degree holders have almost doubled compared to pre-pandemic levels. That increase is still focused on the same information, finance, and professional-services fields that used to be a safe bet for a new graduate. The unemployment gap between young degree holders and their less-educated peers has come to its closest point in 30 years. That’s not at all a sign that things improved for people without degrees. Things actually got worse, and fast, for people who have them.
LinkedIn’s most recent data has entry-level hiring down 6% year over year, with midlevel hiring down 10%. This squeeze is affecting everyone too. As of May 2026, workers with 10 or more years of experience sent almost half of their applications, 49% to be exact, to entry-level jobs according to Indeed’s Hiring Lab. Only 12% of their applications went to senior positions. The people employers claim to want most are stuck applying for the jobs that were designed to go to the people they claim ‘they can’t find.’ Everyone is fishing in the same shrinking pond.
Nobody Has the Time, or Room, to Train
Even where junior jobs do survive, the capacity to actually develop someone in them is starting to thin out. More than 90% of organizations with a manager development program say their own managers don’t have time to participate in them. That’s from a September 2025 report from the Association for Talent Development. Gallup found the same story from three different angles in early 2025. Forty-one percent of employees name time as the biggest barrier to learning and development, and 37% of managers say the same about developing anyone else. Eighty-nine percent of CHROs call time away from the daily grind the single biggest obstacle of all.
Bureau of Labor Statistics data backs this up from the manager’s side. A manager’s on-the-job training averaged 68 days in 2018, and had dipped to 50 days by 2023. Companies who were looking to cut costs have spent years thinning out middle management. And if middle managers are the people actually responsible for training junior staff day to day. When you cut that layer out, you don’t just optimize your org chart. You remove the actual process that used to turn a new hire into someone useful.
Buying Instead of Building
The last piece of the puzzle is what happens to people already on-staff. Internal hiring fell 8% year over year, per Workday’s 2025 Global Workforce Report, even as overall hiring demand rose 6% over that same stretch in the first half of 2025. Promotions dropped in 10 of the 11 industries Workday tracks. Manufacturing is the only exception. Companies are still hiring… they’re mostly just not hiring from inside.
Put the whole thing together and it looks like this. Less internships produce less people with any real-world exposure. The entry-level jobs that are still around ask for experience beginners simply can’t have. Managers who are left standing don’t have the time or the staff beneath them to fix the problem with training. The employees already inside the building, who might otherwise be developed into the experienced hires everyone wants, are getting promoted much less often than they were a year ago. Every employer who chooses to reach outside for someone “already experienced” is drawing from a depleting pool, one that every other employer is also refusing to refill.
Why It’s Rational… and Ruinous for All Employers
There’s an old, well-tested idea in labor economics that explains why this keeps happening, even though almost nobody wants the outcome.
The Poaching Problem
Economist Gary Becker described this back in 1964. Employers are systematically under-investing in training workers for general transferable skills. Why bother, when a competitor can hire that worker away once they’re trained, and capture the return without ever paying the training cost? Economists call this a poaching externality. A recent NBER working paper confirms the effect is still real and measurable today. It’s definitely not just a theory from a 60-year-old textbook. Firms genuinely under-invest in training because of exactly this fear. When public programs manage to absorb that risk instead, firms train more and grow faster.
In plain terms: Training a junior employee well enough that they become genuinely valuable is a risky bet that a rival company won’t just pay them a little more once they’re trained enough to be useful. To avoid the risk, companies just choose to skip the training and just hire someone who’s already valuable instead. Honestly, that’s a perfectly rational move for any single employer to make. They do have to accept that when it’s multiplied across entire industries, it guarantees they will have far less trained people to hire in five years.
IBM’s CHRO Made the Same Case
IBM’s own chief human resources officer, Nickle LaMoreaux, made almost the identical case in public when she announced the company would triple its U.S. entry-level hiring in 2026. Slashing early-career recruitment might save money in the short run, she said. But it risks creating a scarcity of mid-level talent later. That scarcity forces companies to poach from competitors instead, which tends to cost more than promoting from within. Those external hires also typically take longer to get up to speed than someone trained in-house from day one. A Fortune 50 CHRO independently arrived at the same conclusion that a 1964 economics paper predicted. It’s a pretty strong signal that the system is real and needs some serious attention.
Call this what it is: a collective action problem. It’s not: a mystery or a generational complaint about ‘kids these days.’ Every employer who skips training is passively wrecking the resources that all of them will need in a few years… The supply of people who know how to do the job.
AI Isn’t The Only Explanation
The easiest explanation for all of this is that “AI swallowed up the entry-level jobs.” It’s an easy story to sell, and it fits in with the current times. But it’s lacking. The most rigorous recent research on this points somewhere else entirely first.
The Remote-Work Finding
The New York Fed’s Liberty Street Economics blog tried to actually separate the causes. The study came out in June 2026, by research economist Natalia Emanuel, Emma Harrington of the University of Virginia, and Amanda Pallais of Harvard. What they found is that remote work explains about 64% of the recent rise in unemployment among young college graduates, not AI exposure.
Their argument is about proximity instead of technology. People get more feedback and more informal mentorship when they actually work near their colleagues. That feedback goes away fast when people are separated, even by a short distance. The effect hits hardest on younger workers. They have the most to learn and the least standing to ask for help long-distance. The researchers also had access to detailed data from an unnamed Fortune 500 company. While its offices were closed during the pandemic, the firm hired fewer inexperienced workers and more experienced ones. The theory was experienced hires need less mentorship to do the job well. Once its offices reopened, the company shifted back toward hiring younger workers again. But even after reopening, it kept favoring experienced hires for any team that still had remote workers.
That single finding also helps explain something that looks like a contradiction. Especially if you only follow the AI headlines. Around the same time, one survey put the portion of companies cutting entry-level hiring because of AI at 66%. A separate poll of more than 350 CEOs found 67% saying AI was increasing entry-level headcount at their companies. Same quarter, totally different scenarios. The AI narrative is currently kind of arguing with itself. The remote-work research at least gives us a metric to trace and to measure. That always beats a story that can flip depending on which survey you happen to read.
What AI Research Actually Shows
This is not to say AI is blameless, it’s not. Stanford’s Digital Economy Lab published a thorough study in November 2025, using real payroll data from ADP covering millions of workers. It found employment for 22-to-25-year-olds fell specifically in occupations most exposed to generative AI. In jobs where AI is used mainly to expand instead of replace junior work, employment stayed steady or even grew. The authors nicknamed young workers in exposed jobs “canaries in the coal mine.” Researchers went back in February 2026 to check whether interest rates, rather than AI, might explain the pattern. The timing didn’t fit. AI exposure still held up as the best explanation for that specific piece of the data.
A separate Harvard working paper from November 2025 describes something similar using a different name. Seyed Hosseini Maasoum and Guy Lichtinger call it “seniority-biased technological change.” For companies that adopted generative AI, entry-level hiring fell greatly while senior employment at those same companies actually kept growing. Two independent research teams, using different data and different methods, landed on the same type of result.
Proximity Was Never a Problem Before
The honest read is: The consequences of remote work and AI are real, for sure. Remote work looks like the bigger single-driver, so far. But AI has its own separate, measurable effect on top of it. Both problems point toward the same underlying failure. The naturally informal way people used to pick up expertise was by actually sitting physically near someone more experienced and taking over the lower-level parts of the job. That natural pick-it-up-as-you-go part is falling apart from every direction. Erik Brynjolfsson, one of the Stanford study’s authors, put it plainly:
“I think that we’ll have to more explicitly train people, as opposed to just hoping that they will figure these things out on their own.”
Because nobody built any sort of replacement for the training that used to happen automatically, just by proximity.
Who Gets Locked Out Before They Reach the Ladder
Every piece of the puzzle described so far is getting harder for everyone. But, it’s not getting harder equally.
Who Can Afford the Unpaid Middle
First-generation students in California are three times less likely than their peers from continuing-generations to complete a paid internship, according to a 2024 California Competes study. The same structure is seen nationally in NACE’s research. Unpaid internships disproportionately shut out students from expensive metro areas and lower-income families. Those students simply cannot afford to work for free for a whole summer. National data backs this up too. Using numbers from the 2023 Baccalaureate & Beyond survey, the Journal of Student Financial Aid found low-income students much more likely to end up in unpaid internships than high-income students. Attending a highly selective institution, meanwhile, was the single strongest predictor of landing a paid internship.
Handshake’s survey data adds a specific, individual detail to that pattern. First-generation students who do land an internship are more likely than their continuing-generation peers to be juggling coursework or a separate paid job at the same time. It’s 80% versus 70%. They’re not choosing to spread themselves thinner. They just don’t have the option not to.
Luckily, policymakers are starting to not treat this like unfortunate side effect, and more of a two-tier system. California’s AB 323, signed into law October 2025, lets community college districts use workforce development money to directly support paid, work-based learning. The reasoning was clear: Unpaid internships were locking out all the students who couldn’t afford to work for free.
The Network You’re Assumed to Already Have
Money is really only half of it. A lot of internships never get advertised enough to need an application at all. First-generation students have measurably less access to the professional networks and informal hiring channels most employers rely on to fill internship slots. Access matters more and more every year, as the number of open slots shrinks and employers lean harder on referrals to sort through the flood of applications. Knowing the right person does more work than a formal application process used to do.
Geography adds another filter. NACE’s most recent survey put the share of employers offering relocation assistance to interns at roughly half, meaning the other half don’t. That restricts the position to students who already live near where the internships are, or can cover the cost of moving somewhere else for a summer.
This is the same pipeline problem from earlier, just measured even further upstream. When the number of steps on a ladder shrinks, the people who get bumped off first aren’t random. They’re the people who had the least chance to begin with. Taking an unpaid or underpaid year to build the experience companies want is simply unworkable.
The Employers Betting the Other Way
Not every company is playing this game the same way, and those exceptions are worth seriously mentioning, because they actually contradict their industries.
IBM’s Bet
IBM announced in February 2026 that it would triple its U.S. entry-level hiring for the year, across all departments. The announcement came at the same summit where much of the industry was still sharing AI-layoff headlines. LaMoreaux was direct about the reasoning:
“The companies three to five years from now that are going to be the most successful are those companies that doubled down on entry-level hiring in this environment. Not the companies that held the status quo or reduced it.”
She wasn’t pretending AI hasn’t changed the work. She said outright that AI can now do most of what an entry-level job looked like just two or three years ago. IBM’s response was to redesign those jobs instead of cutting them. Junior software developers now spend less time on routine coding and more time working directly with customers and reviewing AI output. Entry-level HR staff step in specifically where the company’s HR chatbots fall short, correcting them and flagging problems instead of answering every question themselves, from scratch.
McKinsey’s Redesign
McKinsey made a similar bet and even released the numbers behind it. Eric Kutcher, senior partner and chair of McKinsey North America, spoke at the firm’s media day in September 2025. The firm plans to grow its North American headcount by 12% in 2026 compared to 2025, he said. Its base of 5,000 to 7,000 non-partner staff could grow 15 to 20 percent over five years. The firm has also changed how it screens it’s candidates. Heather Stefanski, McKinsey’s chief learning and development officer, described a gamified assessment called Solve that tests critical thinking, decision-making, and systems thinking. The firm decided that skill set was becoming a stronger signal than prior business knowledge for who could actually do the job well.
The picture even at McKinsey isn’t perfectly clear. The same “seniority-biased technological change” research found entry-level postings across law, consulting, and finance falling by nearly a third since 2023, even at firms that are growing their overall headcount. So don’t read IBM and McKinsey as proof that the pessimists were wrong. Take them as evidence that a handful of large, well-resourced employers made a specific, public bet: Under-investing in junior talent is a short-term win that leads to a long-term loss. They’re willing to say so on the record, while most of the industry just keeps doing the opposite.
This Has Happened Before
American hospitals have lived through some version of this exact loop about once a generation since the 1930s. Nursing has always trained new nurses through something close to an apprenticeship model, with experienced nurses teaching newer ones on the floor. Every time that system wore down, whether it was through burnout, retirements, or funding cuts to nursing schools, a shortage followed a few years later, and recruiting harder couldn’t fixed it.
The current nursing cycle is a clean picture of this. Nursing school enrollment declined in 2021 for the first time since 2000. The reason happened to be capacity, not interest. Schools simply didn’t have enough faculty or clinical placement slots to teach the people who wanted in. Usually experienced nurses teach nursing students, but they’re that same group who’s now aging out and retiring, in record numbers to boot. The shortage of experienced people is limiting the process that’s supposed to supply more experienced people. It’s just deepening the same shortage everyone is trying to solve one piece at a time. It’s the same loop in a different industry, just running on a much longer timeline.
The lesson from a century of nursing shortages isn’t really complicated… even if fixing it is. You can’t recruit your way out of a shortage if the actual system for turning beginners into experienced workers is what’s broken. You have to rebuild that system directly. Nobody solves a training-capacity problem by adding more applicants to a pipeline that has nowhere to put them.
The Real Question
Becker wrote about this in 1964. The New York Fed measured it in 2026. Hospitals lived through their own version of it starting all the way back in the 1930s. Different decades, different industries, same exact failure. Training someone is a bet, and any single employer can rationally choose not to make it. When every employer makes that same choice at the same time, it ruins the supply of experienced workers a few years out. The people squeezed out of the pipeline earliest always tend to be the ones who had the least room to take on an unpaid or underpaid developmental year. A handful of large employers have already run the math differently, and they did it publicly.
So, who is this economy actually counting on to produce its experienced workers five or ten years from now? Right now, it’s much less companies than the moment calls for, never mind what it requires. And the people paying the cost of this broken system are disproportionately the ones who had the least chance to begin with: no savings to cover an unpaid year, no network, no relocation budget to chase the few real internships that still exist. None of this alone will fix the whole system. That gets decided one hiring plan at a time, and they can still be decided differently.
