Cake, AI, and the Cliff

Gregor Thuswaldner, the provost at La Roche University, posted on LinkedIn last week that he had been selected for the inaugural higher ed cohort of the Google GSV Ed Leader Fellowship. La Roche is where I taught for most of my academic career, and it is exactly the kind of small private institution I have been thinking about for the past year as I have built out work focused on transforming struggling regional colleges. So I went looking at the program. What I found is worth writing about.

The fellowship pairs 27 senior higher ed leaders — provosts, CIOs, vice provosts, academic innovators — with a year-long incubator to design "Transformational Opportunities" at their institutions, using Gemini and the rest of Google's AI stack.

It is a well-run program. The K-12 version has been operating for three years and has touched 5.9 million students. The new higher ed cohort is impressive on paper. I have nothing bad to say about the people in it.

But look at where they sit. Notre Dame. Berkeley. Georgetown. UT Austin. BYU. Florida State. University of Houston. University of Central Florida. UT San Antonio. These are flagship institutions with deep endowments, stable enrollment, and the operating slack to absorb new technology without breaking a sweat. Scattered among them, a different cluster: Lynchburg. Roanoke College. La Roche. Small privates operating much closer to the edge.

The thesis driving the fellowship is that AI, deployed strategically by senior leaders, can be the lever that lifts higher ed into its next era. I want that to be true. I don't think it is.

AI Is a Layer Cake

AI sits on top of the existing institutional model — the courses, the credit hour, the residential campus, the tuition discount, the academic calendar, the org chart. Done well, it makes everything underneath a little better. Faculty get hours back. Students get faster feedback. Advising gets more personalized. These are real gains. I'm not dismissing them.

But for institutions that are already doing well, the layer cake just makes them a slightly nicer cake. Notre Dame with Gemini in the classroom is still Notre Dame. Berkeley with an AI assessment platform is still Berkeley. They will publish the case studies, give the conference talks, sell the dean's report. None of it changes who they are or what they do. It is frosting on something already structurally sound.

For institutions in genuine trouble, the layer cake doesn't help — because there is no cake underneath.

Lynchburg Is the Case to Watch

University of Lynchburg is down to 1,425 students, roughly half its peak. It cut 17 academic programs in May 2024 and laid off 40 staff. It is offering faculty buyouts as of last November. It is on accreditor warning status, partly for financial management. It is, in its own president's words, "at a crossroads."

Two of Lynchburg's administrators are now Google GSV Fellows. One is the Chief Educational Technology and AI Officer — a role that almost certainly did not exist three years ago. The framing from leadership has been consistent and explicit: AI is central to "building the university of the future."

I wish them well. I genuinely do. But no AI initiative — not faculty upskilling, not adaptive tutoring, not AI-driven advising, not Gemini-assisted curriculum mapping — addresses the actual problem at an institution like Lynchburg. The actual problem is that the four-year residential liberal arts model, at their price point, for their region, against their demographic curve, doesn't pencil out anymore. You cannot AI your way to a different enrollment number when the high schools producing your students are shrinking and the kids who do exist are increasingly skeptical that the degree is worth the cost.

A Transformational Opportunity that adds AI to the existing model is, structurally, a way of avoiding the harder transformation. It buys leadership a year of innovation narrative, a Lookbook entry, a stage at ASU+GSV. It does not buy them a future.

The Third Door

There is a transformation the fellowship doesn't put on the table. It is the option of asking whether the institution should still be a four-year residential liberal arts college at all — or whether the right move is to convert the real estate, the staff, the accreditation, the donor base, and yes, the AI tools, into something else entirely.

A regional lifelong learning hub. A workforce development center. A guild-based commons serving learners from 16 to 76 in their actual community. Something built for the world after the demographic cliff, not in denial of it.

That is not a TO Google and GSV are likely to select for. It is not technology-forward enough. It looks, from the outside, like decline rather than innovation. But for a meaningful fraction of small private institutions in this country, it is the only honest path between two unacceptable options: pretend everything is fine until accreditation lapses, or close.

What to Watch

Watch Lynchburg. Watch what their fellows publish at the showcase next spring. Watch whether their accreditor extends warning status. Watch fall 2026 enrollment.

If the AI-first transformation stabilizes them, I will be the first to say I was wrong. I don't think it will. I think what we are watching is a generation of well-intentioned higher ed leaders being handed a beautiful frosting kit and told it counts as building a new cake.

It doesn't. And the colleges that most need a new cake are the ones who can least afford to find that out the slow way.