Higher Ed Is Still Muddling Through AI

It has been nearly four years since OpenAI announced ChatGPT in November of 2022, and most universities continue to muddle along with their AI efforts. Many are treating it as a tactical academic integrity issue when it really is about strategic organizational transformation. If done right, AI can fundamentally change how students learn, faculty teach and conduct research, and universities operate.

Based on our 2025 internal survey of UNC professors, 95% of our faculty agree that our students must leave Carolina ready to use AI intelligently and ethically. Overall, while 90% of college students are using AI, 58% of them believe they do not possess sufficient AI knowledge or skills. Roughly half of students say their institution bans or discourages AI use, and a little less than half say their university encourages it.

Many higher education institutions don’t have a roadmap for AI or, if they do, it’s not being implemented effectively. Having led the Provost’s AI Committee for UNC for 3 years, I will sketch one out below, based on what we have done at UNC and what I recommend doing.

The Roadmap – Getting Started

Before getting too far down the path, you need a philosophy, approach, and framework.

At UNC, our philosophy is, “AI should help you think. Not think for you.” That’s a useful shorthand for faculty, staff, and students working with AI—a quick check for whether a given use of the tool is helping you reason or replacing your reasoning.

The approach I found most useful was intentional adoption—approaching AI thoughtfully. Putting this into the context of an institution’s mission, I’d describe it as, “Performing innovative research and preparing the workforce to shape a better future through the intelligent, ethical, and focused adoption of AI.” “Intelligent” means deploying AI only where it makes sense and building AI skills and infrastructure with the future in mind. “Ethical” means ensuring individual use of AI and AI enterprise deployments avoid bias, protect privacy, and guarantee fairness and accuracy. “Focused” means aligning AI with key missions of the university, thoughtful experimentation, and combining top-down enablement with grassroots, organic participation.

By “framework,” I mean how to think about applying AI to the core functions of the university: teaching, research, and operations. Below those, it’s important to build a firm foundation of AI enablers—free/secure frontier models, AI training, AI governance—that cut across all three.

Teaching

As our faculty research told us, it’s essential that students leave the university ready to work in an AI world. This means two things.

1. Students need AI Literacy—knowing how to personally use AI intelligently and effectively. This goes beyond basic prompting to include understanding how AI works, potential issues like bias, how to validate its output, and how to use it as a thinking partner.
2. They need AI Fluency—grasping how AI will be used in their discipline. Business students need to know how AI is used in marketing, medical students how it’s used for diagnoses, pharma students how it’s used in drug development, and even archaeology students how it’s being used to make new discoveries.

Personal AI literacy should be mandatory in general education, preferably in the first semester, so students also learn the ethical use of AI in academics. AI fluency in their discipline should happen within the major. This requires coordination at the university level across schools to ensure the right topics land in the right part of the curriculum.

Faculty are essential here too. They need training on using AI intelligently and ethically, and they need to redesign assignments to make them AI-resistant.

When our faculty subcommittee considered mandating at least one AI assignment per class, they pushed back hard—faculty wanted the freedom to set their own AI approach per course. I think that makes sense. As long as students leave the university AI literate and fluent via a specific set of courses, not every class needs an AI component. In fact, many liberal arts courses may be better off without one.

Research

For research universities, I think there are four priorities. The first is to offer researchers training on how to ethically use AI in their research. This must include, but should go considerably beyond, ethics. . For example, AI can replace human survey respondents with synthetic research respondents that replicate human responses. This can lower the cost of the research, dramatically speed up results, and improve response rates. 

Second, academia should encourage researchers to study AI’s effects on humanity, society, and individuals. The effect of AI on us as a species and as humans is not only tremendous, but also poorly understood. No other institution is as well-suited to do this as a research university. Areas such as AI’s impact on jobs, social interaction, bias, human meaning, and many others are all ripe for research. Universities can do this with an interdisciplinary flavor if they make it a priority. 

I believe the last two priorities apply only to certain universities. If a university owns proprietary data, that can be a source of unique research findings and leverage. Lastly, if the institution has leading AI researchers, this could be another arena for differentiation (most leading frontier model research will likely be done by the big AI firms like OpenAI, Google, and Anthropic, so universities should focus their research in areas those firms won’t focus).

Operations

Universities are being forced to be more frugal, and AI can help them be more efficient and effective. There are two opportunities here: training staff to be AI literate and fluent in their own roles, and deploying enterprise AI applications—admissions chatbots, HR candidate screening—at the institutional level. Both should be driven centrally rather than by individual schools, so resources are allocated well and best practices transfer across units.

Governance

To execute across a large university organization, there should be a cross-university AI Committee with representation from every school and key departments like IT, much like we had at UNC. It will work in tandem with the Chief AI Officer by providing bottom-up input and execution. Lastly, there needs to be an AI Governance Committee to oversee enterprise AI applications to ensure they are implemented in a fair and ethical manner. Our research told us many faculty and staff still want training in AI Literacy and Fluency, plus a further layer: how to use AI in the classroom, design AI-resistant assignments, and apply AI in their research.

It’s also not enough to offer training—people need access to leading models (Anthropic, OpenAI, Google) free of charge. Because so much of the data being worked on is private or protected by law, these systems must be secure, with data staying within the university’s walls.

One area many universities are still struggling with is policy for acceptable AI use, particularly by students. After all, academic freedom lets each faculty member decide what is allowed in their course, meaning a student taking five courses deals with five different sets of guidelines. We decided to develop ethical guidelines rather than a policy: for students (via syllabi guidance), for faculty teaching and research, and for staff work. We chose “guidelines” purposely, because they’re recommended best practices rather than a mandate, and because AI is changing faster than policy can adapt.

We also built a central AI website—guidelines, updates, ways to get involved—and ran regular surveys of faculty, staff, and student AI usage and attitudes. Together, this information offers the data necessary to set adoption goals, refine strategy, and address community concerns as they arise.

We also identified internal AI enthusiasts to feature on the website and built an AI Community of roughly 1,000 people where the AI-curious can learn and improve their skills. Every university likely has both groups already—the experts and the early adopters—waiting to be found and connected.

Doing the work is necessary but insufficient—people need to be kept informed. Communication about programs, successes, and progress, both internally and externally, is crucial: explain the university’s overall approach, reinforce key messages, tell people how to get involved, and showcase intelligent, ethical AI use.

These steps can’t be done in serial fashion—given how fast AI is moving, they need to happen in parallel across the university. That’s why selecting a Chief AI Officer who knows the institution, has both strategic and execution skills, and is given real power and funding is essential.

While many institutions have used the time since ChatGPT to focus more on academic integrity than strategic transformation, there is still time left on the clock. The roadmap above provides a practical path to get there. It’s now up to universities to take the next steps.

Mark McNeilly is Professor of the Practice at the Kenan-Flagler Business School at the University of North Carolina at Chapel Hill, where he teaches marketing, organizational behavior, and AI. He chaired the UNC Provost’s AI Committee and writes on AI governance and strategy at markmcneilly.substack.com. He is also an Advisor for QuantHub, an AI Edtech firm.