Speaking to attendees at the ASU+GSV Summit AI Show on April 6, Kadiu said students increasingly expect intelligent, personalized experiences from their institutions that traditional staffing simply can’t provide. Kadiu cited author and journalist Jeff Selingo in pointing out that students can wait two to six weeks for appointments with mental health services and up to two and a half hours for financial aid assistance, but they want those experiences to be efficient and personal.
“This avalanche is coming in, and more support is needed,” Kadiu said. “AI can really bridge that gap.”
In 2012, he said, AI was about perceiving the world around us, through functions such as speech recognition and image identification, with common applications in translation and medical imaging. The next big jump was in late 2022 with generative AI, which creates content rather than just analyzing it. In 2025, Kadiu said, agentic AI takes action rather than only responding to real-time prompts to generate content.
That means there are new opportunities to use AI not just as a tool, but as a teammate, and Kadiu predicted that early adopters will have a competitive advantage. Through live demonstrations, he showed his audience a variety of uses for AI agents in higher education, from admissions counseling to campaign strategizing.
Kadiu's company, Element451, is one of many offering AI agent services for higher education. On the surface, some of these services may resemble generative AI systems that users are more familiar with, but behind the scenes, they involve a system of AI agents working together, similar to how human staff would.
Kadiu said AI agents are specialized with a skill type, a goal, a communication style and a set of actions it can take, like initiating an email or phone call. While the agent can take action on its own, it also has a set of “approvers” linked to it who OK each action before it takes place.
For example, a recruiting agent might have the overall goal of getting a student to apply to a university. It will use existing university data systems to determine the best time for outreach, then create an email, SMS message, phone call or WhatsApp message. If that action is approved, the agent will store a transcript of the outreach plus any questions it fields in the student’s file, which the agent will use to inform human staff as well as its next course of action.
“We can just give the goal and provide all the tools and let the AI reason about it, and because it's built on all of the context and the data and all the prior conversations, it's probably the best way to kind of bring all those pieces together and not miss anything, right?” Kadiu said. “Imagine that you have to go from person to person, or that you have to see 500 students. It’s impossible to keep all their context in your head.”
In the live demonstration, when Element451’s Vice President of Product Eric Range told an agent named Carolyn that he was ready to apply to the fictional Element University’s nursing program, Carolyn sent him a link to the application in real time.
Another use case might be for financial aid help, especially with recent Free Application for Federal Student Aid (FAFSA) hiccups. Kadiu said the Element451 financial aid agent is trained on studentaid.gov resources, and institutions can input their own resources as well. Specializing the agents means that one chatbot doesn’t need to know everything about the university, it just needs to know what the other agents do.
“They can hand things off to each other, just like a real human team would on the support channel,” Kadiu said.
For example, when Range used a chatbot to ask about scholarships, a financial aid agent named Finley answered his questions. When he moved on to questions about athletic scholarships and asked how to try out for the soccer team, Finley passed him to a different agent specializing in athletics named Coach Carter.
Outside of community-facing communications, the reasoning abilities of AI agents can make them helpful for things like application management. Kadiu showed off an agent that evaluated applications based on a given criteria and gave a preliminary score.
“It’s making references to the documents: ‘Shows a GPA of 3.7, which places them above the 3.5 threshold. However, none of the provided documents include evidence of at least three AP honors courses,’ and that was part of the criteria,” Kadiu said.
The agent also assessed the application for fraud.
In the future, Kadiu said, AI agents could be added to course management systems to provide personalized interventions for students based on criteria like not attending class or missing assignments. But there are limits to what he will ask AI agents to do.
“We made a clear delineation in the sand, so to speak, around, ‘Hey, we don’t want to mess around with the class pedagogy right now,’” he said.
