The Door Is Closing Before They Even Get There

Last week I marched in a graduation procession and watched students cross a stage after years of work, sacrifice, and in many cases real financial risk. It was a good day. And it also left me with questions I've been sitting with ever since.

Jeffrey Sonnenfeld and his team at the Yale School of Management just published a piece in Fortune that reframes the AI jobs conversation in a way worth paying attention to. The argument isn't that AI is causing mass layoffs. It's something more subtle.

What if the biggest impact of AI on jobs isn't the layoffs we can see, but the opportunities that quietly never materialize in the first place?

What the Numbers Are Showing

Unemployment among recent college graduates has climbed to nearly 6%, rising twice as fast as the rest of the workforce since 2022. Job market confidence has dropped from roughly 70% of workers saying it was a good time to find a job in 2022, down to 28% now. And in a pattern worth examining, college graduates are now more pessimistic about the job market than people without degrees. Only 19% of college graduates say it's a good time to find a quality job, compared to 35% of those without degrees.

That's an interesting reversal. The credential that was supposed to open doors is now associated with more anxiety, not less. Why?

According to Sonnenfeld's research, it's not primarily mass firings. It's a quiet freeze. Companies are getting more output from the same workforce as AI handles more of the routine work. So the entry-level roles that used to be the first rung aren't being eliminated so much as they're simply not being created. There's a difference between being laid off and never being hired, and right now it's the second one that seems to be happening more quietly and at scale.

What Higher Education Is Grappling With

In a survey conducted ahead of the Yale Higher Education Summit in January, 100 college and university presidents were asked about AI preparedness. Only 10% said their graduates were sufficiently or very well prepared for AI-enabled workplaces. Nearly a third said graduates were somewhat or fully unprepared.

That's worth sitting with. Higher education leaders are privately acknowledging a gap between what we're preparing students for and what they're walking into. That's an honest and uncomfortable thing to admit, and it raises a real question: what do we do about it?

I work at a law school. The most important thing we give students isn't information. It's the capacity to think, reason, and exercise judgment in complex situations. That capacity isn't easily automated. But the entry-level roles that used to let young lawyers build that capacity over time — the first jobs, the first cases, the first chances to figure things out in lower-stakes environments — those are among the roles most exposed to AI-driven workflow changes right now. And you can't develop judgment without the chance to use it.

Sonnenfeld's team makes a point I keep returning to: becoming AI-savvy isn't a product of a course or a credential. It's a mindset built through experience. And experience requires access. If entry-level work is becoming harder to find, what happens to the formation that was supposed to come from it?

Who Has the Least Runway

This question matters differently depending on who you are. Students with networks, family resources, and the ability to be patient in a slow market have options that others don't. First-generation graduates who took on debt because a degree was supposed to be the path forward are in a different position. Students who were told to outperform and outwork their way to opportunity are now navigating a market where the opportunity itself has become harder to find.

I'm not saying the market is broken beyond repair or that a degree no longer matters. I genuinely don't know how this plays out. What I do think is worth asking is whether higher education is accounting for the students with the least margin for a delayed start, and whether our support structures are built for the market that actually exists right now rather than the one we graduated into.

Questions Worth Asking

Sonnenfeld's research notes that employers are increasingly looking for people who can exercise reasoning in AI-enabled environments, not just execute tasks. That's a real shift in what's valued. The question is whether we're helping students develop that capacity, or whether we're still optimizing for the skills that mattered most a decade ago.

I watched those students cross the stage last week and felt all of it at once. The pride in what they'd accomplished. The genuine uncertainty about what comes next. The hope that the work they put in will open the doors it was supposed to open.

I hope it does. And I think the least we owe them is an honest conversation about what we're seeing, rather than reassurance that everything will work out fine.