Who Should Be Driving Your University’s AI Policy? (Hint: It’s Not the IT Department)

There’s a particular kind of knowledge that you can only get by doing something over and over again until you understand it from the inside, and teaching is one of those things. You don’t learn what students actually need from a policy document or a technology report or a vendor demo. You learn it at 11 o’clock on a Tuesday night when you’re sitting at your kitchen table, looking at the assignment you just handed back, realizing that twenty-three out of thirty students missed the same point, and asking yourself, “what did I do wrong?” That kind of knowledge is hard-won and deeply specific, and it lives in the people who are still doing the work, year after year, semester after semester.

So when I look at what’s happening at universities across the country right now, with artificial intelligence being rolled out at a pace that would make your head spin, and I see that the people doing most of the driving are IT departments and senior administrators, I get a little worried. Not because I think IT professionals are bad at their jobs. They’re not. They’re usually very good at them. The issue is that their job is not teaching, and right now, the decisions being made about AI in higher education are teaching decisions dressed up as technology decisions.

Here’s what I mean by that. When an IT department implements a technology system, they’re thinking about things like security, scalability, compatibility, and cost. Those are all legitimate and important concerns, and I wouldn’t want anyone cutting corners on them. But what they’re often not thinking about, because it’s genuinely outside their frame of reference, is what it feels like to be a student in a writing class who suddenly has to navigate questions about academic integrity that their professor can barely answer, or what it means for a graphic design instructor to figure out how AI image generation changes what’s worth teaching and what isn’t. Those are curriculum questions. Those are classroom questions. And the people with the most relevant expertise to answer them are the faculty.

The data on this is striking, and we now have enough of it to move beyond anecdote. In July 2025, the American Association of University Professors published a report titled Artificial Intelligence and Academic Professions, based on a survey of five hundred AAUP members conducted in December 2024. Ninety percent of respondents said their institutions are already integrating AI into teaching and research. And yet seventy-one percent said that administrators “overwhelmingly” lead those conversations, gathering what the report described as “little meaningful input” from faculty members, staff, or students. A separate finding from Inside Higher Ed’s own 2024 survey of chief academic officers found that only twenty percent of colleges and universities had published any policy at all governing the use of AI. We are, in other words, moving very fast in a direction most institutions haven’t bothered to write down.

What makes this even more complicated is that a lot of faculty don’t even know what they’re already using. The AAUP survey found that while only fifteen percent of respondents said their institution mandates the use of AI, eighty-one percent said they are required to use educational technology platforms like Canvas and Google Suite. The catch is that AI-powered analytics are now embedded in both of those systems, running quietly in the background even when users have turned the AI features off. Faculty are interacting with AI every day, often without knowing it, because the people who deployed those platforms didn’t think it was necessary to explain that detail to the people actually teaching with them.

And the results of that approach are showing up in the numbers. The same AAUP survey found that seventy-six percent of faculty respondents said AI is deflating their job enthusiasm, sixty-nine percent said it is hurting student success, and sixty-two percent said it has created worse outcomes in the teaching environment. Forty percent said it is eroding academic freedom. Those aren’t the numbers you get when a technology rollout goes well. Britt Paris, co-author of the AAUP report and associate professor of library and information science at Rutgers University, put it plainly in an interview with Inside Higher Ed: “AI in higher education is barely even functional and tech companies view higher education as a cash cow to exploit.”

Here’s the part that doesn’t make it into the reports, though, and it’s the part I think matters most if you actually want to understand why this problem persists. Faculty know what their students need. They’ve always known. The reason they’re not in the room when these decisions get made isn’t because no one thought to invite them. It’s because most of them have already learned, through hard experience, that raising your hand in an institution carries a cost. You speak up about what a new AI tool is missing pedagogically, and suddenly you’re on the implementation committee, unpaid, on top of everything else you’re already doing. You share an approach that’s working in your classroom, and six months later it comes back to you as a mandatory training module that looks nothing like what you described, built by people who were never in your classroom and don’t understand why you built it the way you did. You push back on a rollout that isn’t ready, and you spend the next semester fielding quiet questions about whether you’re a “team player.” And so most faculty do what people do when the system keeps taxing them for participation: they put their heads down, they take care of their students as best they can within whatever constraints they’ve been handed, and they keep quiet. Not because they don’t have the answers. Because they’ve learned it costs too much to give them.

The economics underneath all of this are worth naming plainly, because they don’t get named often enough. According to the AAUP’s own data snapshot on contingency in U.S. higher education, sixty-eight percent of all faculty now hold contingent appointments, meaning positions without tenure and without the institutional standing tenure provides, up from forty-seven percent in 1987. The AAUP defines contingent appointments broadly, including both full-time positions that aren’t on a tenure track and the part-time, per-course adjunct positions that make up roughly forty percent of the faculty workforce on their own, according to CUPA-HR’s 2026 report on adjunct faculty. Only thirty-two percent of faculty hold tenured or tenure-track positions today. The people doing the largest share of instruction, the adjuncts hired course by course, earn a median of thirty-four hundred and ninety-eight dollars for a typical three-credit course, according to CUPA-HR’s 2024–25 Faculty in Higher Education Survey. An adjunct teaching what would amount to a full load (four courses each semester, thirty-six credit hours across the year) would earn approximately forty-one thousand nine hundred and seventy-six dollars annually at that median rate, and the vast majority teach far fewer courses than that, since seventy-nine percent of adjuncts teach six or fewer credit hours per term. Meanwhile, according to Educause research cited by multiple higher-education compensation analysts, university chief information officers typically earn between one hundred ten thousand and two hundred ten thousand dollars per year, with senior CIOs at larger research institutions earning considerably more. McKnight Associates, an HR consulting firm specializing in higher education compensation, reported in August 2025 that presidents, provosts, and senior administrators were seeing annual pay increases in the six to eight percent range, while the College and University Professional Association for Human Resources found that tenure-track faculty received the lowest raise of all five employee categories for the fourth consecutive year, with real salaries sitting eleven point seven percent below pre-pandemic levels when adjusted for inflation. When the person making the call about which AI platform to adopt earns that kind of salary, and the person who has to live with that call every day in a classroom earns a fraction of it, you have a structural problem, not a communication problem. Telling faculty to speak up more, to advocate for themselves, to get a seat at the table, is a little like telling someone to speak up at a meeting they weren’t invited to, in a building they don’t have a key to, on a schedule that doesn’t account for the other work they’re doing to make ends meet.

Marc Watkins, director of the AI Institute for Teachers and assistant director of academic innovation at the University of Mississippi, named the threat directly in the same Inside Higher Ed report: “A lot of faculty are aware that if you start letting a technology like AI dictate the material conditions of your work, you can then have that technology, administrator or state legislature decide to pay you more, less or assign you more work on top of it. Or potentially replace us.” That’s not a paranoid reading of the situation. That’s a clear-eyed one, and the data backs it up.

I want to be clear that I’m not blaming the IT professionals who are doing these jobs, because that would be missing the point entirely. They’re doing what they’re trained and hired to do, and most of them are doing it well. The problem is systemic, not personal. It’s that the architecture of how universities make decisions about technology was built without ever seriously asking whose expertise matters most when the technology in question is about teaching and learning. And now we’re trying to figure out something as consequential as artificial intelligence inside that same architecture, and acting surprised when the results reflect who was and wasn’t in the room.

There’s actually a growing body of formal advocacy that speaks directly to what would need to change. The AAUP’s AI report draws on the organization’s own 1966 Statement on Government of Colleges and Universities, which established that it is “the responsibility primarily of the faculty to determine the appropriate curriculum and procedures of student instruction,” and makes the case that this responsibility has to extend to AI and educational technology decisions as well. The report calls for faculty-led shared governance, mandatory impact assessments before any new technology is deployed, and genuine protections for faculty who have sound pedagogical reasons to decline specific tools. Separately, in a document first published in 2024 and revised in March 2025, a group of researchers and practitioners published an AI Bill of Rights for Educators through the EngageAI Institute, a framework developed primarily for K-12 settings but increasingly cited in higher education discussions, which explicitly lists among its core principles the right to “agency in pedagogy,” defined as the ability to make choices about “whether, when, and how to use AI based on learning goals, student populations and learning contexts, and pedagogical judgment.” That’s not a radical position. That’s just asking the people who understand the classroom to have real authority over what goes into it, not advisory access that can be overruled the moment the answer is inconvenient.

Think about it this way. If you were redesigning the way a hospital delivers patient care, you would absolutely want the technology team at the table, and you’d want to hear what they had to say. But you wouldn’t put them in charge of clinical decisions. You’d put clinicians in charge of clinical decisions, because they’re the ones who understand what good patient care actually requires, and you’d ask the technology team to support that vision, not define it. The principle is exactly the same in a university. The difference is that in a hospital, the authority structure generally reflects whose expertise matters most for patient outcomes. In higher education, it often doesn’t, and the gap between who holds the knowledge and who holds the power to act on it has been widening for a long time. AI is just making it harder to look away from.

The tools being deployed today are just the beginning, and the governance structures we build now, or fail to build, will shape everything that comes after. The faculty who have spent decades learning how students learn, how to design an assignment that requires real thinking, how to build a course that holds together from the first week to the last, those people have something that no technology platform and no administrative committee can manufacture. And a lot of them are sitting quietly in their offices right now, on a Saturday night, planning next week’s lessons, with plenty of things they could say about what their students actually need from AI and what they don’t. They’re just not saying it, because the institution has taught them, over and over, that it’s safer not to.

That’s the problem. And no amount of well-intentioned policy language is going to fix it until the institutions doing the deploying are willing to honestly ask themselves who’s being heard, who isn’t, and why.