Gen AI graduates to operations in higher ed

ChatGPT-written term papers? That’s so last semester. Since the AI chatbot’s 2022 debut, CIOs at the nearly 4,000 US institutions of higher education have had their hands full charting strategy and practices for the use of generative AI among students and professors, according to research by the National Center for Education Statistics. But now higher ed CIOs are beginning to turn their focus to using gen AI to improve operations.

Universities are increasingly leveraging LLM-based tools to automate complex administrative processes. They’re handling student applications, financial aid, resource allocation, faculty workload balancing, and compliance reporting as well as back-office functions like procurement. Many universities also find gen AI-powered chatbots can provide students answers to a lot of pressing questions and concerns, from fees to anything revolving around campus life.

The potential savings are enticing, too. Just ask Ravi Pendse, VP for IT and CIO at the University of Michigan. One of the earliest proponents on gen AI use for learning, Pendse discovered the technology’s value for operations when the university’s internal billing department replaced a legacy procurement tool that cost hundreds of thousands of dollars. Using an AI tool built on the university’s Maizey LLM dropped the annual cost to just $62.

Even better, it can be changed easily. “This is a modular system,” says Pendse. “AI components can be swapped out, making it flexible and cost-efficient.”

The Maizey gen AI platform is also available to the university community, which can use its point-and-click interface to easily create apps. Plus, it empowers subject matter experts to address their own needs.

“We say, ‘Here are the tools. Here’s how they work. Now go take this and figure out where it’s most helpful to your community,’” he adds. U of M has used Maizey to create a travel portal where chatbots answer common staff and professor questions, greatly reducing calls and emails staff used to have to handle manually. These FAQ-type apps appear to be a major win.

“A chatbot is perfect for that,” Pendse says. “Historically, the folks who manage travel have heard every possible question. Now, staff can focus on more urgent matters.

Human oversight still critical

Not long ago, a lot of universities penalized students for using AI. Now they’re adopting ways of integrating AI into their operations, says Dhriti Banerji, a project manager and founder at Higher Edge consultancy. Banerji helps Australian universities like RMIT scale up projects like chatbots and virtual assistants to handle student inquiries. Many of her current projects focus on freeing up staff to focus on personal interactions, which are especially important in the university setting, such as where a student needs extra support to stay in school.

Banerji’s clients have also used gen AI to collect enrollment data to do timetabling for class scheduling, engage facility capacity, and handle staff rostering. As a result, such tools significantly cut down on process times.

“Timetabling used to be very complicated,” she says. “When students enroll in a course, it needs to align with other students’ schedules, curriculum needs, and staff availability. Previously, it was a highly manual process that took a lot of time. But now, AI tools help generate these schedules though human intervention is still there to verify everything aligns properly. We’re just too new at this to completely depend on AI.”

One of Banerji’s clients is using gen AI to streamline the research grant-writing process, which essentially involves filling out an application. “AI can reference previous grants, suggest improvements, and help researchers complete applications in a shorter period of time,” she says. And having an LLM act as co-pilot could ensure no stone is left unturned when identifying potential research grants as a tight funding climate takes hold.

Start with student pain points

When it came to using gen AI to improve university operations, Lev Gonick, enterprise CIO at Arizona State University, started with two perennial student concerns. Right after receiving that coveted acceptance letter, a student’s joy often turns to anguish as they or their parents embarks to conquer financial aid forms.

“Financial aid is a bear,” says Gonick. “We’ve been focusing on all kinds of technologies to streamline and create timely feedback for students and tools they can use for self service. But don’t expect these functions to transition to gen AI overnight.”

Last year, there was about a 37% uptake of students self-serving in the financial aid app, and he expects usage will increase over time as students enjoy the convenience of using it.

The second? Parking. “We’ve created an AI agent called Parky in honor of our Sparky mascot,” says Gonick. “It’s fully automated and can transact any business related to parking on campus.”

Gonick and his team are working with software vendors including Salesforce and ServiceNow to make sure their offerings align with what’s most pressing at ASU. “We’re not waiting for vendors to come sell us something,” he says. “We’re telling them what we need based on what students need.”

For example, gen AI capabilities embedded within ASU’s ServiceNow IT service management implementation provide a streamlined experience for many aspects of student operations, including setting up and making changes to their health and technology services. “We do millions of ServiceNow tickets every year,” he says.

ASU also keeps an open door policy for gen AI and LLM tools, rather than standardize on a few. “Right now, we support 55 large language models,” says Gonick. “We’ve worked with OpenAI, Google, and will make some other announcements at our ASU GSV Summit in April.”

Quality control

Sunay Prasole, assistant vice chancellor for engineering remote education at Texas A&M, says he sees a lot of opportunity for how the university will use gen AI in operations. But much remains aspirational right now.

Texas A&M was looking to use gen AI agents to pinpoint prospective students and he asks, “What if we could have an AI agent that mines student data and says, ‘People like you have been admitted to Texas A&M or UT in the last five years. Would you like to apply?’” The AI agent finds prospective students, fills out the application for them, and submits it on their behalf.

Prasole’s team ran a pilot gen AI admissions project, but testing immediately identified a problem. They took all the data for one year’s admitted students and ran it through the gen AI to see whom the model would select. “It should hopefully identify the same people for admission,” he says. “But it excluded a bunch of people who were admitted.” They’re now working to fix the issue, he adds.

In the near future, Prasole expects to use agentic AI for a series of specialized tasks such as resource allocation in enrollment management. But regardless of how it’s adopted, the university won’t use gen AI to cut staff. “It’s not about eliminating jobs,” he says. “It’s about allowing people to focus on the human connections that matter most.”

Working to strengths

University operations is a natural environment where bots handle time-consuming manual processes while enabling people to do what they do best.

“AI can try to imitate people, but it can never replace true empathy,” says Pendse. The University of Michigan uses AI in student advising to answer basic questions, but when they have higher-level concerns, they go directly to a human advisor. “The AI tool reduces wait times, gets students instant answers, and allows advisors to focus on students who need deeper support, he adds.

AI could also be useful across the board to help identify students who need some high-touch interaction, such as when someone misses a class three times in a row. In that case, a five-minute empathetic phone call from a professor can do more for student retention than any AI tool ever could.

AI isn’t a choice anymore, as Gonick puts it, and Pendse believes gen AI will be the most transformative technology of the century. “It can be a force for positive disruption if we use it thoughtfully, ethically, legally, and with contributions from all aspects of our community.”