Artificial intelligence has the potential to transform higher education by connecting student experiences, improving operations, and enhancing outcomes. But unlocking its full potential will require better data integration, deeper cross-campus collaboration, well-defined ROI measures, increased environmental transparency, and more nuanced conversations about its role.
In November 2022, ChatGPT 3.5 was released to the public, ushering in a new era of artificial intelligence (AI) for the world and higher education. The past thirty-six months have been filled with curiosity, concern, excitement, and anxiety. In higher education, we have seen powerful new tools and services alongside growing worries about jobs, learning, creativity, the environment, and intellectual property.
Three years in, I want to offer a few reflections on the journey so far, challenges that remain, and the promises that continue to emerge.
Enabling a Longitudinal View: Unlocking the Full Potential of AI
One of the most powerful aspects of generative AI is its ability to learn about the user over time. The more people engage with tools such as ChatGPT, Copilot, or Gemini, the better they become at drawing on past interactions to deliver more personalized responses and insights.
Our students do not engage with us in a vacuum. Their interactions flow across the entire institution, from the dining hall to the classroom, the gym, the library, and into the residence halls. These discrete moments combine to form each student's unique journey.
AI offers a way to connect those moments. It can integrate the totality of a student's experiences across the institution to help inform decisions, anticipate needs, and enhance both engagement and success.
For example, an AI system might know that Sally asked about the rugby team when she applied, that she later declared a major in English, that she's in the honors program, and that she submitted a facilities report when her residence hall room flooded last semester. Connecting these pieces of information can help the institution better anticipate her needs, tailor communications, and offer more holistic support—such as surfacing information about upcoming rugby matches, alerting advisors to potential stressors, or recommending writing-intensive courses that fit her interests.
The power of AI in this context lies not in any single data point, but in its ability to recognize patterns and context over time. This longitudinal view, when guided by human judgment, ethical boundaries, and respect for privacy, can transform the student experience from reactive to anticipatory, and from transactional to deeply personal.
The problem is that much of the marketplace—and many campuses—are not architected to leverage AI in this way.
- Point solutions limit the broader potential of AI. Today's marketplace is largely comprised of point solutions. Every day, my inbox fills with messages from vendors touting AI tools designed to improve specific campus services. Yet these tools rarely communicate with one another or have the capability to access and learn from a person's experiences beyond the siloed confines of their particular function.
- Data must be architected for a longitudinal view. We must ensure that our data repositories are designed to support a secure, comprehensive record of each student. The key to using AI to enhance student success, elevate the student experience, and improve operational effectiveness lies in enabling AI tools to securely and appropriately access all relevant institutional data.
- Institutional silos hinder progress. To achieve a holistic view of the student experience, we must work across traditional institutional silos. While this is not primarily a technology issue, building bridges between departments and campus service providers is essential. Doing so ensures that everyone understands how their work contributes to the holistic student journey and the impact they have along the way.
While each of these tools may offer meaningful benefits on its own, their isolation limits the broader potential of AI. Without the ability to connect insights across the wide range of student experiences, we miss the opportunity to create more significant and lasting impacts for students.
Effective data governance and a robust data environment—such as a centralized data lake or warehouse—are more essential than ever. The widespread move to software-as-a-service (SaaS) solutions over the past decade has made access to some key datasets more complex. Yet as AI capabilities advance, the ability to connect and analyze institutional data in secure, ethical, and integrated ways has become even more critical to realizing its full potential.
Working across silos will inevitably reveal both gaps and strengths in institutional processes. It will surface disconnects where coordination is lacking and highlight examples of excellence worth emulating. Meeting this moment will require deeper conversations and sustained collaboration across campuses.
Articulating Return on Investment
As higher education institutions move from experimentation and pilots to the production and scaling of AI tools across campus, higher education leaders need to be able to articulate the return on investment (ROI) of these efforts. Even before the popularization of ChatGPT 3.5, we were using AI tools to deepen our understanding, improve operations, and enhance our work. Yet quantifying those benefits ahead of time can be difficult.
In numerous conversations with CIOs and other institutional leaders, I've heard a common refrain: "I know AI is providing value, but I struggle to clearly articulate the ROI at the outset of projects." Much of this challenge stems from the fact that AI introduces new tools, techniques, and approaches, leaving institutions without established reference points and metrics to calculate ROI before a project launches and real-world data emerges.
ROI for AI projects typically falls into three broad categories: increasing revenue, decreasing expenses, and making strategic investments to strengthen institutional capacity. Revenue gains often come from improved enrollment and retention enabled by predictive analytics and early alert systems that help identify and support students more quickly. Expense reductions usually arise from automation and operational efficiencies, such as streamlining administrative workflows, optimizing energy use, or improving IT monitoring. Strategic investments, meanwhile, include building data and analytics infrastructure, developing AI literacy among faculty and staff, and piloting innovative applications that enhance teaching, learning, and the student experience.
More work is needed to better identify the potential ROI at the outset and develop tools and metrics that allow us to assess success once initiatives are underway. As AI becomes more embedded in institutional operations, the ability to define, measure, and communicate value will be critical to sustaining momentum and investment.
Engaging in More Nuanced Conversations About AI
Praises and concerns about AI are voiced daily, but too often they're framed in overly broad terms that flatten what, in reality, is a deeply layered set of issues.
Even after three years, many well-intentioned discussions paint AI with a single brushstroke, overlooking the continuum of types and uses that exist. We need to find ways to engage in richer, more nuanced conversations that weigh the distinct benefits and concerns of different uses and approaches. The use of generative AI in the arts, for example, raises very different questions than the use of AI to identify a student who is at risk of not completing their degree. Each use case carries its own implications and deserves to be examined on its own terms rather than being grouped under a single, catch-all narrative about AI.
On our campuses, we need to continue engaging in conversations about AI and find ways to move beyond the surface. Hosting forums and discussions focused on specific use cases or issues, such as the environmental impact of AI or its application in supporting student success, enables deeper exploration and leads to better-informed decisions. The field of AI is vast, and our approaches to understanding it must reach beyond the headlines.
Increasing Transparency About the Environmental Impact of AI
Across our campuses, concerns persist about the environmental impact of AI. How much energy does a single ChatGPT query consume? Is it more or less than an application programming interface (API) call to a large language model (LLM) from an application we develop? How can we architect solutions to reduce overall energy use?
Right now, we lack the tools and metrics needed to clearly understand the energy consumption of different AI solutions. Without this information, it's difficult to make informed choices about design, deployment, and usage. We also need this data to engage our campus communities in meaningful, nuanced, and informed conversations about responsible AI use and to address the legitimate concerns being raised about its impact. Establishing a clear, standardized method for reporting or rating the energy consumption of AI tools, such as an indicator of energy use per query or transaction, would help institutions make more sustainable and transparent choices.
AI solution providers play a critical role in addressing this gap. They must provide greater transparency and more usable data on energy consumption so that institutions can make responsible decisions about how they architect AI-powered systems and when it is appropriate to use specific tools or alternatives.
Achieving Big Impacts Without Big Investments
The past three years have shown that AI can deliver meaningful results without major investment, especially when institutions build on existing infrastructure and expertise. For example, my institution (Ithaca College) leveraged its merged IT and analytics organization and prior investments in analytics infrastructure to develop an AI tool that helps identify and support students who show signs of distress. Our studies have shown that when meaningful interventions occur, retention among these students increases. In its first year, our AI tool contributed to improving student retention, protecting institutional revenue, and, most importantly, helping students continue their college journey.
The tool was built in-house using a small amount of Python code and OpenAI APIs, with data accessed securely through our institutional data lakehouse. Initial development took about eighty hours, and ongoing costs are about twenty-five dollars per month. In its first year, it enabled over 150 additional student interventions—a significant outcome for such a modest investment.
As we in the higher education community adopt AI tools to support our institutions, we should continue to build upon existing systems, architectures, and expertise. We also need to design our infrastructures with AI applications in mind. While these solutions may not always match the sophistication of commercial products, simple, targeted approaches can often deliver meaningful, high-impact results.
Preparing Students for an AI-Powered Workplace
While operational investments are crucial, our greatest responsibility lies in equipping students to thrive in a rapidly changing professional landscape. Much has been written over the past eighteen months about the impact of AI on entry-level jobs.Footnote1 As the job market continues to shift, and many entry-level positions are replaced or supplemented by AI tools, higher education institutions must continue to evolve their curricula accordingly. Students need a grounding in how AI is used in their fields of study, including the skills required to use AI effectively, and a deep understanding of the ethical, privacy, and societal considerations that accompany its use.
Equally important, students must be able to clearly articulate their "human intelligence"—the skills, experiences, and value they bring to the workplace. The ability to work collaboratively, think critically, communicate effectively, exercise judgment, and lead with empathy become even more essential as AI reshapes the early stages of many professions.
Preparing students for an AI-powered workplace, therefore, requires both the ability to use AI responsibly and skillfully and a strong foundation of human intelligence that distinguishes their contributions. Together, these capacities will enable graduates to navigate and succeed in the emerging world of work, and our institutions must continue to help students develop and strengthen these skills.
Building on Lessons Learned: The Next Three Years
The last three years have been exciting, and the next three hold great promise as we apply what we've learned to expand the scale and positive influence of AI tools while engaging in important conversations about their broader implications. By addressing the challenges outlined above and continuing to foster substantive, informed dialogue about AI on our campuses and across higher education, we can build on the progress of the past three years and ensure future work strengthens our institutions, supports our people, and advances our mission.
Notes
- Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen, "Canaries in the Coal Mine? Six Facts About the Recent Employment Effects of Artificial Intelligence,"working paper (Stanford Digital Economy Lab, Stanford University, November 13, 2025). Jump back to footnote 1 in the text.
David Weil is Senior Vice President, Strategic Services and Initiatives, and CIO at Ithaca College.
© 2025 David Weil. The content of this work is licensed under a Creative Commons BY-NC-ND 4.0 International License