What AI Can—and Can’t—Do in Higher Ed Enrollment

While many higher ed enrollment leaders are now eager to use AI for predictive modeling, communication, and student engagement, questions about the potential ethical considerations—and about how AI actually enhances recruitment and retention—may continue to create obstacles among stakeholders seeking consensus regarding new policies and strategies.

To help explain AI’s promise and potential role in enrollment management, Ashton Black, Manager of Data Science for Liaison's Othot AI solution, recently answered several important questions about AI in practice, the role of human judgement when using the technology, and how institutions can leverage data to support student success and enrollment goals responsibly.

What are the most effective real-world uses of AI you’ve seen in enrollment management so far?

Ashton Black (AB): Some of the most effective applications of AI in enrollment management come from predictive modeling that helps institutions better understand and anticipate student behavior across the lifecycle—from prospect to graduation. For example, models that interpret the likelihood of enrollment, retention, persistence, and graduation enable institutions to identify which students are most likely to matriculate and succeed, and where additional support or outreach may be needed.

Other models can further enrich this understanding, such as predicting the likelihood of achieving a GPA above a critical threshold (e.g., avoiding academic probation) or utilizing holistic models that help institutions align their evaluation process with mission-fit priorities by identifying applicants who most closely reflect the values and characteristics they’ve defined as important.

These models are most impactful when grounded in strong, comprehensive data inputs—including academic performance, engagement activity, financial context, and demographic background—that together create a fuller picture of each student. The resulting predictions give enrollment teams actionable insight into where to focus their limited time and resources for the greatest effect.

What truly makes these models powerful, however, is transparency. Incorporating explainable AI techniques—such as feature impact values—reveals why a prediction is what it is and how much each factor contributes. For instance, a school may learn that financial aid awards or campus visits have a stronger influence on enrollment for certain students. That level of visibility builds confidence in the model outputs and ensures that staff can interpret and act on them appropriately.

The next layer is prescriptive analytics, which helps institutions move from understanding what is happening to knowing what to do next. By surfacing the most effective actions or interventions for specific students or groups, enrollment teams can prioritize outreach and tailor communications strategically. When combined with segmentation and aggregation of individual insights, these approaches allow leaders to identify broader institutional patterns, assess the effectiveness of strategies, and allocate resources more efficiently—creating a continuous feedback loop that enhances both student engagement and strategic decision-making.

Where do you believe human oversight is still critical in the student recruitment and admissions process?

AB: Human oversight remains essential for interpreting and contextualizing the aspects of a student’s experience that data alone can’t capture. While AI models are excellent at identifying patterns and probabilities, they lack the nuance to understand personal conversations, motivations, or life circumstances that can profoundly influence a student’s decisions, especially when those factors aren’t captured in data, or are too unique to be meaningful within a model or the broader population.

For example, a counselor might know that a student is very likely to enroll because of a personal connection—such as having a close friend already attending the institution—even if the data doesn’t reflect that relationship. Similarly, an advisor might learn that a current student at risk of leaving is struggling because of a sick family member or a sudden change in financial status. These insights, though not always quantifiable, are critical to interpreting predictive results responsibly.

AI is most effective when it works alongside human judgment—automating what can be quantified while empowering people to apply professional expertise, intuition, and empathy where the data ends. This ensures that models are used as guides, not as absolutes, and that compassion and individualized understanding remain central to the decision-making process. The combination of human and machine intelligence ultimately leads to better, fairer, and more student-centered outcomes.

How can institutions adopt AI responsibly without over-relying on automation or losing the personal touch?

AB: AI should be seen as a tool to enhance human decision-making, not replace it. One of the most effective ways institutions can adopt AI responsibly is by using it to summarize and surface key insights that would otherwise be impossible to process manually, given the vast amount of data institutions manage and the limited time and resources available to staff. Rather than automating decisions, AI can highlight patterns, identify priorities, and provide a clear, data-informed foundation for human interpretation and action.

For example, by segmenting student populations based on predictive scores—such as likelihood to enroll or retain—institutions can better allocate their team’s time and effort.

How do the insights provided by AI help institutional leaders address the unique needs of individual students?

AB: Students with lower probabilities of retaining may require more immediate and personalized attention, while those with higher scores might simply need light-touch engagement or monitoring. This approach ensures that staff focus first on the populations where their involvement can make the greatest difference, maximizing both impact and efficiency.

Importantly, these scores should never be viewed as the final answer but rather as a starting point—a signal to look more closely. Human context, such as a student’s motivation, life challenges, or expressed intent, remains critical for interpreting the full picture accurately.

Beyond informing individual-level understanding, model insights can also spark broader conversations about strategy to help institutions decide which actions to take, at what scale, and within their available resources or risk tolerance. When used this way, AI doesn’t diminish the personal touch—it strengthens it—by giving staff the insight and time to engage more meaningfully with the students who need it most, while supporting a holistic, mission-aligned approach to student success.