Addressing an audience of such leaders at the 2025 EDUCAUSE annual conference in Nashville this week, Managing Director Alexander Brown of technology consulting firm Attain Partners said as a baseline, each project should align with the institution’s mission and consider the different levels of trust among various user groups, their capacity for training and education, and infrastructural capacity to accelerate other projects in line with the rapid pace of emerging technology.
To help frame that work, Brown suggested IT leaders begin designing institutional AI strategy with three questions, covering project size, focus and governance.
1. Should your institution start with small, targeted AI initiatives or big, sweeping investments?
Some institutions are testing small, contained projects while others are offering campus- or even systemwide access to AI tools.
Szymon Machajewski, academic technology leader at the University of Illinois, Chicago (UIC), said shifts in the tech market have made it less cost-prohibitive for colleges to go for campuswide AI readiness. For example, he said UIC has invested in the AI writing tool Grammarly for students, faculty and staff.
“Large businesses like Google are now ready to be the partners, and they make it really great,” he said. “So, it's time to do something big that is not just individual pieces.”
Dave Comroe, CIO of Teachers College at Columbia University, offered a different perspective.
“I’m not sure what ‘go big’ looks like other than lighting a bunch of money on fire,” he said.
To him, making a big investment in something so new without testing it at a small scale doesn’t add up.
Marylou O’Donnell-Rundlett, a leader in enrollment and student administration at Boston University (BU), said it is possible to do both. BU has made its chatbot, TerrierGPT, available to all students and employees while running smaller pilots with enrollment services to manage email and call volumes, she said.
2. Should your first AI project look outward to serve the student population, or inward to assist faculty and staff?
Institutions also differ on whether to begin with student-facing improvements or with internal efficiencies that could free up staff capacity.
O’Donnell-Rundlett said operational gains are especially important with limited human resources.
“For us to be able to do the student experience we need, I need to free up some people,” she said.
Rich Janes, a program director for technology resources at the University of Texas at Austin, said he has the same goal but thought starting with student-facing tools could reach it, as well.
“Provide [students] some things so that they don’t need the staff quite as much, and then free [staff] up to do other things,” he said.
Jason Martin, an IT director at the University of North Carolina's pharmacy school, said the question involved balancing safety and governance. With the varied responsibilities of administrators, he said it might be easier to create a governance structure for student-facing projects.
“Are we going to make grading better? Are we going to make student organizations better?” he said. “If it’s grading, certainly [consider] FERPA, or for enrollment, Social Security numbers. But that's much easier to draw a box around, in my opinion, than all of the duties that might fall under administrators.”
3. Should your AI governance be centralized under IT or distributed across colleges and programs?
With colleges and departments’ differing needs, the panelists agreed those communities should play a role in driving AI strategy that suits them.
Comroe views his role in IT leadership as more focused on providing appropriate tools and resources than providing centralized governance.
“What are the needs? What do you want to do with this, departments?” he said. “Tell me, and collaboratively, we’ll make this happen.”
Machajewski said that because AI is such a unique technology, it calls for distributed management more than other IT operations do.
“We see that in pedagogy — liberal schools will grade differently than engineering, and engineering may want AI all the way whereas liberal schools are saying 'analog,'” he said. “That's why I think the distribution [method] is a little bit better.”
Janes said that AI governance could be centralized, but not under IT, because it is a tool like the Internet. He said having IT lead the transition to the Internet at UT Austin, without enough input from business and other departments, created disconnect.
“In our team having dozens, if not hundreds, of conversations around AI over the last year, even for just these really simple, basic starting questions, there's not a universal answer,” he said. “We may have had universal answers around how we think about rolling out technology for other systems that we’re more familiar with.”
