In September 2026, a committee of 18 Cornell faculty published a 238-page report on the future of the American university.¹ It makes nearly 70 recommendations.² Its central diagnosis concerns what it calls the transactional approach to college, in which families pay for a credential and the university supplies it. The committee’s verdict on that approach is blunt: “It optimizes for signals of competence at the direct expense of competence itself.”¹

The report sets this against a transformative education aimed at lasting capability. “The transactional approach provides a signal of learning,” it says, “while the transformative approach provides the results of learning.” Ariel Avgar of Cornell’s ILR School, one of the committee’s four co-chairs, told The Boston Globe: “The relationship between students and the university has become quite transactional.” He tied that partly to cost: “Part of the reason for that transactional relationship has to do with the high sticker price.”³

The committee’s sharpest evidence came from employers, whose skepticism the report calls “more jarring.” They “consistently reported that American graduates, whatever their specialized knowledge, arrive in the workplace without some of the professional capacities the workplace demands.”¹ The capacities employers report finding less reliably in new graduates are “the ability to handle uncertainty and solve problems that do not have clear answers, a readiness to absorb critical feedback and recover from setbacks, and working in teams that require compromise with others who do not share their perspective or assumptions.”

Harvard’s dean of undergraduate education, Amanda Claybaugh, documented the climb in an October 2025 review of grading (Exhibit 1).⁴ The review concluded that “our grading no longer performs its primary functions and is undermining our academic mission.” Admissions committees, it reported, say “it can be difficult to tell Harvard students apart.” One faculty member called the situation “a farce.”

Harvard’s faculty voted 458 to 201 in May 2026 to limit A’s to 20% of each course’s undergraduates, plus four, starting in fall 2027.⁶ Yale’s Committee on Trust in Higher Education reached the same diagnosis in April 2026: “Decades of inflation and compression have rendered the college grading system almost meaningless as an academic measure.”⁷ It recommends a mean grade of 3.0 and course percentiles on transcripts.

The most cited evidence comes from sociologists Richard Arum and Josipa Roksa. They tested more than 2,300 students at 24 institutions with the Collegiate Learning Assessment, which measures critical thinking, complex reasoning and writing.¹¹ Forty-five percent showed no significant improvement in those skills during their first two years. After four years, 36% still showed none, and the average gain was less than half a standard deviation.¹²

Effort was modest too. Students in the study spent about 12 to 14 hours a week studying, and 32% took no course that assigned more than 40 pages of reading a week.¹² The skills the test measured showed up later in the labor market. Graduates who scored in the bottom fifth were unemployed at a rate of 9.6%, against 3.1% for those in the top fifth.¹³

Economist Bryan Caplan argues that most of college’s payoff is signaling. In The Case Against Education, he contends that a degree certifies intelligence, work ethic and conformity more than it builds skill.¹⁴ By his own estimate, “80% of the causal effect is signaling.”¹⁵ His strongest evidence is the sheepskin effect: “Senior year pays far more than the earlier three years combined!”¹⁶

The signal is strongest at the top. Fewer than one in a hundred college attendees go to one of the 12 Ivy-Plus colleges, yet their graduates fill a far larger share of the country’s most prominent positions (Exhibit 2).¹⁷

Part of that pattern is selection, since these colleges admit students who were already likely to succeed. Raj Chetty, David Deming and John Friedman estimated the effect of attendance itself. Compared with a typical flagship public university, an Ivy-Plus education raises the odds of three elite outcomes:¹⁸

The average graduate sees much less. The same paper finds “no significant effect on actual mean income rank at age 33.”¹⁸ The credential changes who reaches the very top, and much of that runs through the employers who recruit on these campuses.

Sociologist Lauren Rivera studied how those employers choose. She interviewed 120 people who hire for elite banks, consulting firms and law firms.¹⁹ Educational prestige decided who got an interview. After that, “fit was typically given more weight than grades, coursework, or work experience even in first-round interviews,” and more than half of the interviewers ranked fit as their most important criterion. One banker summed it up: “A lot of this job is attitude, not aptitude.”

AI may now be wearing down the signals themselves. Luis Garicano, an economist at the London School of Economics and a co-author of the 2026 book Messy Jobs: The Work That AI Cannot Reach, lists hiring among the processes firms will have to rebuild. “You cannot do a very good screening,” he said in September 2026, “because all the CVs and all the signals are worthless.”²⁰

Each year the National Association of Colleges and Employers asks employers which career-readiness competencies they consider essential, and how proficient they find recent graduates in each. For its Job Outlook 2025 report, 237 employers answered both questions.²¹ Their answers locate the gap precisely (Exhibit 3).

The gap is not new. In NACE’s survey for Job Outlook 2018, 99.2% of employers called critical thinking essential and 55.8% rated recent graduates proficient in it. For oral and written communication, the figures were 95.9% and 41.6%.²²

Graduates see themselves very differently. NACE’s companion survey asked 2,281 graduating seniors to rate their own proficiency in the same eight competencies (Exhibit 4).²¹

The pattern held in NACE’s spring 2026 update. Employers there were “less impressed with new grads’ communication, professionalism, and critical thinking skills,” and 22% called new graduates very well prepared.²³

The questions differ from survey to survey, so the figures are not directly comparable. The direction never changes: the people closest to the teaching are the most convinced it worked.

The World Economic Forum’s Future of Jobs Report 2025 surveyed more than 1,000 employers across 55 economies.²⁷ Analytical thinking led their list of core skills, followed by resilience, flexibility and agility, and by leadership and social influence (Exhibit 5).

Strada Education Foundation and the Burning Glass Institute traced graduates’ careers through online career histories and Census data. Their study counts a graduate as underemployed when the job does not typically require a degree.²⁸

The New York Fed’s measure points the same way. In mid-2026, 42% of recent graduates aged 22 to 27 were underemployed, against 34% of all college graduates.²⁹

Goldman Sachs is the sharpest case. More than 360,000 students applied for its 2025 summer internship.³⁰ In 2026 it took 2,500 interns, fewer than one in 100 applicants, for a third straight year below one percent.³¹ Yet the firm’s president, John Waldron, says the qualities that will matter most in an AI economy are not the ones a transcript shows.

Asked in September 2026 how AI changes Goldman’s recruiting, Waldron said the firm did not yet know the answer. “I think our thesis would be thinking and analytical problem solving and intellectual curiosity,” he said. “Old-fashioned elements like that are gonna be back in vogue.” Goldman, he added, is “more focused on what’s the well-roundedness of your background and have you proven that you actually are on a little bit of an intellectual journey.”³²

Waldron’s own path makes the point. Goldman first turned him down: “I didn’t do well in the financial modeling test.”³² Days after the podcast he wrote that the firm attracts graduates “who understand numbers and logic—but we also look for people who have the communication skills and EQ needed to build strong relationships with clients and across the firm.” His advice to students: “start your day by reading.”³³

He is candid about what this generation finds hard. “Most people are better talkers than listeners,” he told students at Brigham Young University. Young people now grow up with “more information overload than I had when I was your age, so it’s going to be even harder to assimilate all that information.”³⁴ He also sees a strength: “You are not as afraid, I observe, to step out.”

Goldman’s chairman and CEO, David Solomon, frames the AI problem as one of apprenticeship. Firms must now teach young hires “a variety of things that they’re not going to intuitively learn, because they don’t actually have to work as hard to get the answer.”³⁵ Intelligence alone will not carry them: “the smartest person in the world without a whole package of other things, is not going to navigate Goldman Sachs well.” The rest of the package, in his words, is “the human elements, the ability to connect, resilience, determination, striving for excellence.”³⁶

Jacqueline Arthur, Goldman’s head of human capital management, names the scarce part: “what’s much harder to instill are qualities like judgment, adaptability, and critical thinking.”³¹

JPMorgan Chase’s chairman and CEO, Jamie Dimon, has gone further. “For a lot of jobs, we got rid of college degree required,” he said in 2024. “Skills are far more important than the college degree.” He added: “I don’t think necessarily because you go to an Ivy League school or have great grades, it means you can be a great worker.”³⁷ In his April 2026 letter to shareholders, he wrote that “high schools and colleges should be judged by their outcomes — do the students graduate with job offers, and what do they pay?”³⁸

His advice to young people is a list of competencies: “critical thinking, learn skills, learn your EQ, learn how to be good in meetings, how to communicate, how to write.”³⁹ He worries that remote work left “younger kids kind of being left behind. They weren’t developing their EQ as much.”⁴⁰

At Bank of America, chair and CEO Brian Moynihan described the odds graduates face. “200,000 applications, we hire 2000 people,” he said. “If you ask them, if they’re scared, they say they are.”⁴¹

The Burning Glass Institute and Harvard Business School tracked 11,300 roles at large firms from 2014 to 2023.⁴⁹ The number of roles that dropped a degree requirement rose almost fourfold. The hiring results barely moved:

Degree requirements are also returning. Indeed found that 19.3% of U.S. job postings required a bachelor’s degree in November 2025, up from 16.6% two years earlier and close to pre-pandemic levels.⁵⁰ A few large employers have moved further. Walmart says 90% of its U.S. roles do not require a degree,⁵¹ and about 40% of Bank of America’s hires do not have one.⁵²

For most graduates, the credential still gets them considered. What the leaders quoted here are saying is that it no longer decides who is hired, or who advances.

Wharton psychologist Adam Grant put the evidence plainly in The New York Times: “Academic excellence is not a strong predictor of career excellence. Across industries, research shows that the correlation between grades and job performance is modest in the first year after college and trivial within a handful of years.”⁵³

The research behind that claim is a meta-analysis that pooled 71 studies covering nearly 14,000 people.⁵⁴ Its central finding is how quickly the relationship decays (Exhibit 7).

Google reached the same conclusion from its own hiring data. “G.P.A.’s are worthless as a criteria for hiring, and test scores are worthless — no correlation at all except for brand-new college grads, where there’s a slight correlation,” Laszlo Bock, then its senior vice president of people operations, said in 2013.⁵⁵ Some Google teams were 14% people who had never gone to college.

Grant’s explanation is that grades measure the wrong things. “Academic grades rarely assess qualities like creativity, leadership and teamwork skills, or social, emotional and political intelligence,” he wrote. “Getting straight A’s requires conformity. Having an influential career demands originality.” He cites a study of job postings in which “nearly 15 percent of recruiters actively selected against students with high G.P.A.s.”⁵³

In Hidden Potential, Grant names four character skills: being “proactive, prosocial, disciplined, and determined.”⁵⁶ His central claim is that they can be learned. “I now see character less as a matter of will, and more as a set of skills,” he writes. “Potential is not a matter of where you start, but of how far you travel.”⁵⁷

Cornell’s committee reaches the same place from the employer side: “Workers get poached to new firms because someone admires their work, not their degree.”¹ It also argues that judgment is learned among people. “Sound judgment is interpersonal: it requires the willingness to risk being wrong in front of others, to hold and be held accountable.”

Grant warns against reading any of this as a verdict on young people. “Don’t label an entire generation as lazy because they’re not willing to make the sacrifices you did,” he wrote in January 2026.⁵⁸ Of Gen Z, he has said: “Maybe the work ethic is the same, but there’s lower tolerance for burnout or for indentured servitude.”⁵⁹

For most of the past 35 years, a new college graduate was less likely to be unemployed than the average worker. That edge has disappeared (Exhibit 8). Since January 2021, recent graduates’ unemployment rate has been above the rate for all workers every month, and the annual gap has widened every year since.²⁹

The squeeze is sharpest where AI is strongest. Using payroll records for millions of workers, Stanford’s Digital Economy Lab estimates that employment of 22- to 25-year-olds in the most AI-exposed occupations is 19% below where it would be had it kept pace with less-exposed peers.⁶⁰ Most of the gap comes from reduced hiring. The estimate has grown with each update, from 13% in August 2025 and 16% in November 2025.

Industry data point the same way, with less rigor. SignalFire, a venture firm that analyzes LinkedIn-derived profiles, found new graduates made up 7% of hires at the 15 largest technology companies, with new-graduate hiring down 25% from 2023 and by more than half from 2019.⁶¹ The Burning Glass Institute found unemployment among bachelor’s holders aged 20 to 24 rose from 5.2% in 2018–19 to 6.2% in the two years to June 2025.⁶²

Economists who study how jobs are put together are more measured. “The disruption of junior career ladders is real,” Garicano wrote in April 2026, but the prediction that half of entry-level white-collar jobs will disappear “confuses task automation with the extinction of jobs.”⁷⁰ What changes is the junior job itself. For knowledge workers, MIT Sloan labor economist Anna Stansbury suggested at the September discussion of Garicano’s book, “the job has become a management job even at the entry level.” Machines can now do much of the execution, and doing the execution was how juniors learned to define the question, set the goal and judge the quality of the work.²⁰

Harvard’s grading review quotes a student: “We are terrified of the A-.”⁴ Cornell’s committee heard that students see “grades below an A as an existential threat to their life path,” and that they “often seem checked out of their classes, doing what is required to get the best possible grade, but not engaging with the material for the sake of learning.”¹ In a transactional model, the report says, “curiosity is a liability.”

AI tools add a new temptation. OpenAI’s Sam Altman has observed that college students “don’t really make life decisions without asking ChatGPT what they should do.”⁷² Cornell’s committee puts the institutional version of the question: “What is the value of a university credential if the outputs on which students are assessed may have been produced by AI tools?”¹ Its answer is the cultivation of sound judgment: “This capacity is what will justify a human’s seat at the table.”

The pressure starts before college. In September 2026, PISA, the OECD’s international assessment of 15-year-olds, reported that U.S. reading scores had fallen 14 points since 2022: about six or seven months of learning by the OECD’s estimate, and the same drop as the average across OECD countries. One in three students in PISA’s survey said they used AI to draft text for writing assignments, and the data show “a generally negative correlation” between using AI to aid writing and reading scores, though PISA cannot say which causes which.⁷³ Garicano thinks self-discipline is becoming more valuable. Students now face “a larger opportunity cost of doing the homework,” with TikTok and YouTube competing for the time, and “a lower cost of not doing it, because you can use AI.” He pointed to the new PISA results as a sign that “some of this may be going on.”²⁰

Families are drawing their own conclusions about the degree.

This briefing is not an argument against college. The credential still opens doors, and the most selective degrees open the most. The evidence argues for building competence alongside the credential, and for starting before the stakes are real.

Some of that building has to happen outside the lecture hall. Garicano, a professor at the London School of Economics, put it plainly: universities are “not necessarily set up for teaching people how to do stuff on the job.”²⁰

Six capacities recur across the Cornell report, the employer surveys and the executives quoted here. Each can be practised in high school.