AI in higher education operations means using AI agents to carry out the administrative work that keeps a campus running, such as building course schedules, processing financial aid, and following up on compliance steps, so staff can spend more time with students. Most of the conversation about AI in higher education has focused on the classroom. A smaller group of leaders is asking how AI can change the way the institution itself runs.
That question was the center of a recent episode of Audacious Ideas, a podcast on leadership and decision-making hosted by Dr. Lillian Schumacher, president of Tiffin University. Her guest was Zack Perkins, co-founder and CEO of CollegeVine. This post pulls out the main ideas for campus leaders.
What is the difference between AI for learning and AI for operations?
AI for learning covers how AI is used in teaching: what students may do with it, what faculty decide, and what institutional policy says. Perkins noted that higher education moved early on those questions and that the jury is still out on where they land. AI for operations covers the work of running the institution, and he sees far less activity there.
"The realm, though, of using AI to change our operations is still so, so nascent," Perkins said. In his conversations with university leaders, he added, very few institutions have even 2% of their operations touching AI, beyond individual staff members using ChatGPT as a personal assistant. (For more on how CollegeVine defines the category, see Higher education is asking the wrong question about AI.)
He described what operational AI looks like across a campus:
- A registrar producing better course schedules that account for retention, persistence, and how buildings and faculty are allocated
- A financial aid process that moves faster and awards more equitably
- Finance and business operations, including supply chain and inventory management
Perkins argued that work on AI in the classroom and work on operations should run in parallel. His concern is a university that works out its future learning model and then cannot deploy it. He said "the biggest tragedy to higher education would be if they solve that, they figure out what that model should be, but they can't implement it," because the institution has too much inertia or its legacy software cannot accommodate the change.
Why should leaders focus on enduring problems instead of today's solutions?
Perkins said leaders often pick the wrong problem to solve: one that is too low-level, too local, or one that will not exist in the same form tomorrow. AI shows the pattern clearly. A workflow that once justified its own software tool can disappear when an agent handles the whole job, and the next problem looks different again.
His answer is to anchor on problems that do not change and be willing to swap the solution stack constantly. For a university, he named serving students and the community, keeping strong revenue-generating functions, and staying agile. He keeps CollegeVine's own goal at that level on purpose. "I'm not saying I want this department in this subdivision to solve this one problem," he said, "because that problem is not going to exist a year from now depending on how we refactor that organization entirely."
He also drew a line between operations and mission. "Your operation is not your mission. Your operation serves your mission," Perkins said. A university can be fully mission-driven and still run a tight operation, and he finds that the institutions that fall short of their mission are often the ones that tangle the two together.
What does a student-centered operating model look like?
Perkins started with how a student experiences a university. A student may be in contact with something like 30 departments, with compliance steps they do not know about and dependencies between requirements they cannot see, such as a course taken at the wrong time that adds a semester to graduation. A one-stop office helps, he said, but it is still a lens on top of siloed departments and older processes that may not need to exist.
CollegeVine is working with partners to map the institutional logic behind degree pathways, including credit requirements, and put it into software. With that in place, the questions change. Can the pathway be shortened? Which choke-point courses need more sections? Which compliance steps are getting in the way and could be dropped? (Perkins covered the shift toward this kind of deployment in his Vineyard 2026 keynote.)
Dr. Schumacher described Tiffin University's version of the idea, called Dragon Pathways. Students are placed on a path from the moment recruitment begins, before they reach campus, and are supported through persistence and completion with help customized to them. Tiffin has worked on the model for four years and has seen results in persistence and completion, she said. She described AI as a way to focus more tightly on the parts that move the needle, and the two organizations are exploring how AI can strengthen the model.
Perkins pointed out that guiding a journey does not mean one journey for everyone. The Dragon Pathways design gives students many pathways to choose from, and the structure still helps them avoid mistakes they do not need to make while they explore. He contrasted that with universities that remain an open playground and expect students to direct themselves. AI can also give leaders, in his words, "an amazing data feedback loop" on how pathways are performing and where they can improve.
Can operational AI improve margin without layoffs?
Perkins disputed the projections that say many universities will run out of money. In his view, many of them assume operating costs keep growing at the same pace as in past decades, when part of that growth tracked enrollment growth. With enrollment likely to flatline, he called that assumption faulty.
10 to 15 points of margin. Perkins's estimate of what a university could free up by rethinking its operating model, including where AI can run an office and which functions to centralize.
He said most of that margin would come from somewhere other than headcount. His examples came from what he sees in institutional data:
- A compliance issue that takes an office 40 days to resolve, when the data shows students drop out if it is not handled within two weeks
- Classroom allocation that pushes sections into time slots that shut out adult learners with childcare pickup
- Patterns that look unrelated until the data connects them, such as HVAC sensor data at a larger university showing colder buildings tied to lower attendance, which traced through to lower pass rates and persistence
"There are so many buckets of just odd inefficiency that have nothing to do with headcount," Perkins said. He acknowledged that some jobs will not need to exist, and said he does not expect that to be the majority of the change.
How can AI agents run operations safely?
Perkins described how CollegeVine approaches the risk that AI can make mistakes. The approach connects to a university's existing data systems and builds an operating model from them: the business operations in each department and the rules that govern them. Leaders then get a view of where processes are breaking down and how many students are affected, traced down to the net tuition revenue lost each semester. AI agents can then act on those processes with guardrails.
Policy decisions stay in deterministic logic. An agent does not decide whether a student receives a particular aid package or whether an applicant is admitted. "The agent isn't deciding what package to give the student. It's looking at what packages are allowed to be given," Perkins said. Deterministic software handles that question, and the agent then carries out the step.
Where should leaders start with AI in operations?
Perkins gave two pieces of advice.
The first is to get close to the work. He spends time on sales calls, partner calls, hiring interviews, and shipping product himself, because that is how he learns how things actually run. He suggested senior university leaders do the same: package a financial aid award for a student, sit in on admissions committee reading, or try to build next semester's schedule with the registrar. Do not just ask an AI to tell you what is happening, he said. Inspect it yourself.
The second is to hold two timelines at once. Some leaders start a one-year committee to plan AI before acting, and by the time it finishes, the landscape has changed. Others roll out the first thing they see, which he called dangerous. The leaders he sees doing this well keep the long-term vision and also have "the pragmatism to say let's cut this down into the thing that we can get started with this month," then evaluate it as they go.
He encouraged leaders to treat this period as an opening. What an institution does between 2024 and 2028, he said, is what its success will later be traced back to.
Why is Perkins optimistic about higher education?
He gave three reasons. First, higher education is a bundle: workforce preparation, credentialing, a soft landing between adolescence and adulthood, research, and support for local communities. He noted that many venture-backed alternatives over the past 20 years have not taken hold, and that even if AI handles 80% of tactical work, he expects roughly as many people to attend college as do today. Second, changes to the operating model can make institutions financially sustainable. Third, he does not expect a job apocalypse.
Looking ahead, he predicts a more diverse set of institutions, with smaller ones that do certain things well able to compete. Over the next 10 years, he expects a more fluid learner journey in which people return to the university over time and institutions work more closely with employers. For more examples of where operational AI is already running, see 6 AI use cases gaining traction in higher ed operations.
.png)
.png)
.png)
