Corporate America is having second thoughts about its enthusiastic embrace of AI. This spring, Microsoft canceled most of its licenses for one of the most popular AI coding tools, partly over cost. Uber, meanwhile, burned through its entire 2026 AI budget in four months. After two years of buying AI as fast as they could, companies are starting to ask harder questions: What should this technology actually be for, where is it making us more efficient, and where should we draw a line?
Higher education is no different. Half of colleges have begun using conversational AI tools at an institution-wide scale, and roughly 4 in 5 undergraduates were using AI in their studies as of October 2024, according to a survey of 11,706 students across 15 countries by the student support platform Chegg. As AI becomes more widely used across higher education, colleges are gaining a stronger understanding of where it can be most effective, where a person’s judgment is still essential, and where guardrails are needed to keep the two in balance. A well-designed AI tool can encourage more students to file financial aid forms and register on time. AI can answer questions long after the advising office has closed. In fact, according to one student support platform, more than a third of inquiries to chatbots arrive outside business hours. Human advisers, however, are far better equipped to understand why a student has stopped showing up to class or to help them balance a heavy course load with the rest of their complicated lives.
It’s time for institutions to create a new compact around the ethical application of AI, drawing clear boundaries between what the technology can handle well and what it cannot, and optimizing human connection in the moments that matter most.
Institutions can start by taking a closer look at the administrative functions required to keep students on track. AI easily earns its keep with early-warning alerts, well-timed nudges ahead of deadlines, and answers to routine questions at any hour. Today, advisers must handle much of this administrative work themselves. With the National Student Clearinghouse Research Center reporting that almost 40 percent of students who start college don’t earn a degree within six years, institutions need every hour of adviser time to go where it helps most.
But automation alone cannot solve higher education’s completion challenge. Students dealing with imposter syndrome, basic needs insecurity, loneliness and life disruptions need the kind of emotional intelligence that only a person can provide. When students explain why they consider leaving, they cite emotional stress, mental health, cost and work demands more often than academic concerns. Better-timed outreach will do little to help a student navigate that landscape.
Targeted student support is the best antidote higher education has to the pressures that push students out. At the same time, that kind of sustained personal support in higher education can be cost- and resource-intensive. Coaches, advisers and other support specialists have massive caseloads, averaging nearly 300 students, according to 2021 research from the consulting firm Tyton Partners. Institutions can pair the analytical power of AI with the relational depth of human support. Software can identify the students who most need that attention, while optimizing the coach’s schedule so they can build trust, foster belonging and guide students through the hardest decisions standing between them and a degree.
Before institutions deploy AI as part of their communication infrastructure, it’s paramount that they invest time in understanding which conversations should include humans and which can be automated. An AI tool can remind a student when a financial aid form is due, but a person should be the one asking a student why they stopped showing up to class.
Institutions are already road-testing this kind of deliberate boundary-setting. For example, National University’s Sanford College of Education, California’s largest provider of teaching credentials, recently unveiled a framework that assigns every course one of five levels of AI integration, ranging from actively encouraged use to full prohibition. Rather than leaving each classroom to improvise its own rules, National’s RAISE 5 framework provides a shared standard and starting point for faculty and students.
The State University of New York, meanwhile, recently adopted a systemwide AI policy covering all 64 of its campuses. The policy requires campuses to provide training and establish their own guidelines by the end of this year so that faculty, staff and students can use AI responsibly. This is exactly the kind of guardrail-setting that’s needed, not just for how AI is used in the classroom, but how it is governed across the institution.
Higher education has spent the past two years determining what AI can do. Now, it must confront the tougher challenge of defining explicit roles for AI and for people, while ensuring the handoff between them is seamless. Colleges should lead with human relationships and deliberately decide which conversations AI can handle on its own and which should always have a person at the helm. Those constraints won’t limit the technology but allow higher education to get even greater value from it.
Ruth Bauer is the president of InsideTrack, a student success nonprofit that has provided one-on-one success coaching to more than 3 million college students since its founding in 2001.
Tony Frey is the president of Mainstay, an ed-tech company that creates AI-enabled student support tools that colleges use to improve enrollment, persistence and completion.
Leave a Reply