I've spent much of my career moving between two worlds that rarely trust each other: enterprise software, where the instinct is always to scale, and higher education, where scaling the wrong thing can quietly break what makes an institution work at all.
When I was leading product through BCG's acquisition of MAYA Design, and later at Nielsen, "scale" was often the answer on the table. Bigger reach, more automation, faster iteration: the kneejerk logic of technology companies is to remove friction wherever you find it. For a lot of problems, that logic is useful.
But universities aren't systems of infinite scale. And what we need to understand is that it’s not a flaw to be engineered away; it’s actually the whole point. You can lay a six-lane highway down the center of a campus so students simply drive in, go to school, and drive out… but then you’ve lost the undulating paths through campus, the ease of access to shops and cafes, the manicured grounds for hosting events and hangouts, the impromptu meets around dorm buildings. You’ve lost the human scale, and thus, you’ve lost the school.
The Friction Is the Feature
Ask anyone who's tried to move fast inside a university why it's so hard, and you'll get a familiar list of complaints: shared governance, tenure, committees, consensus-building that takes a semester to produce an authoritative paragraph. From the outside, especially from a Silicon Valley vantage point, this looks like institutional dysfunction. An organization that hasn't been "disrupted" yet.
That reading only holds if you're measuring against the wrong goal.
A university isn't optimized for speed. It's optimized for something much harder to engineer: durability of judgment across generations who will never meet each other, and cohorts who will never meet again. The slow, deliberative processes that frustrate efficiency experts are the same processes that let a 150-year-old institution employ the same brand of discipline today that its establishing body would recognize. That's institutional memory doing its job.
The tension is showing up in the data, not just in campus politics. The Digital Education Council’s 2026 global survey of higher ed students and faculty found that AI use has jumped sharply in a single year, but the guardrails around it haven't kept pace: most students still don't feel their institutions or instructors are equipped to guide them on how to use AI responsibly, and most faculty say they haven't been meaningfully included in shaping the policies governing it. That gap is exactly what "moving fast" produces when the institution isn't built for it.
Human Scale, Not Hyperscale
Human scale isn't about size. A university with 40,000 students can still operate at human scale, and a five-person startup can lose it entirely. Human scale is about whether the systems inside an organization are legible and palpable to the people who depend on them. Can a first-generation student figure out how to register for the right class? Can a faculty member understand why a policy exists? Can a provost trace a decision back to the values that produced it?
AI, deployed carelessly, is exceptionally good at breaking legibility. It can automate a process so thoroughly that no one inside the institution can explain anymore why it works the way it does. The aforementioned Digital Education Council survey found real anxiety about exactly this kind of erosion: a large majority of students worry AI is making learning too shallow and discouraging critical thinking, and faculty concern runs even higher. The real risk isn't that AI replaces people. It's that it quietly erodes the human-scale relationships that make an institution trustworthy in the first place.
Why Hyperscaling Doesn’t Pass The Smell Test In Higher Ed
A slower-paced disruption of higher ed is actually in AI's own best interest, not just the institution's. AI companies that show up expecting universities to behave like enterprise software buyers will burn trust fast, get boxed out of procurement, and set the whole category back. The companies that succeed will be the ones willing to move at the institution's pace and design for its actual constraints from day one: privacy, governance, and human legibility, not bolted on after the fact.
Even EDUCAUSE's 2026 Horizon Report lands in the same place: adopting any AI tool is simultaneously a decision about student data, vendor relationships, equity, and trust, and the report's core warning to institutional leaders is that responsible AI adoption can't be outsourced to vendors, chased reactively, or driven by crisis response. That's not a talking point. That's the sector's own research arm telling its members the same thing I'm telling AI vendors, as both a professor and a leader in the space.
What Ought Not to Scale
Not everything in the higher ed experience should be optimized for throughput. The advising conversation that runs long because a student needed to talk something out. The committee debate that feels inefficient but produces a decision people actually believe in. The professor who remembers a student's name three years after they graduated. None of that scales. All of it matters.
The right question for AI in higher ed isn't "how do we scale everything?" It's "what should we scale, and what should we protect precisely because it doesn't?" Getting that distinction right is, I think, the single most important design decision any AI company serving universities will make.
Universities have spent centuries building institutions that keep pace with human beings, not the other way around. In the AI age, that might be their most valuable export. It's also the design principle I've tried to build into skilly from day one: technology that expands what a university's people can do, without asking the institution to give up the judgment that makes it trustworthy to begin with. If I’m hiking to the pinnacle with higher ed as my partner, you can bet I’m looking to them and saying, “your pace.”