Universities Test AI Agents for Administrative Tasks While Teaching Applications Remain Uncertain
Colleges and universities are piloting agentic AI systems - software that can plan sequences of steps, take action, and adjust based on results - across administrative functions. Yet experts caution that decision-makers need to examine reliability, regulatory compliance, and vendor experience before expanding use into teaching and learning.
The term "AI agent" entered mainstream education conversations around 2023-2024. Unlike earlier generative AI tools that answered questions or generated content, agents can work independently for hours, using databases and learning management systems as tools to complete complex workflows.
Where Agents Are Working
Administrative tasks show the clearest early wins. The Illinois Institute of Technology automated transcript processing in 2023, cutting processing time from roughly a month to a single day. The system handles intake, international grade conversion, and customer resource management integration.
Highline College in Washington state introduced a financial aid status tracker in 2023 that reduced emails, phone calls, and in-person visits about application status by 75 percent.
Inside learning management systems, agents are automating instructor workflows. Instructure's Canvas now includes an agentic tool that follows natural language commands. When an instructor says "grant this student an extension," the agent updates the due date, generates a reminder for the student, and flags the assignment for separate grading.
Advising represents another strong use case. AI agents can generate optimized course schedules, test different degree pathways, then route those options to human advisers and schedule follow-up meetings.
Teaching and Learning Presents Obstacles
The conversation shifts when agents enter the classroom. Experts worry that automation could undermine what educators call "productive struggle" - the difficulty students experience when learning something new.
The Einstein agent, created by startup Companion, integrates directly into Canvas and completes assignments automatically. Researchers say this raises fundamental questions about why students should struggle through assignments themselves.
Jake Burley, a researcher at the Applied Ethics Center at the University of Massachusetts, Boston, said education differs from routine institutional tasks. "There's a strong sense that there's something personal or powerful about the educational experience," he said.
Some institutions are finding middle ground. One California instructor created a course-specific AI tutor using custom GPTs trained on their own learning materials. The University of Luxembourg adopted an AWS framework that uses agents across the instructional cycle - from lecture preparation to real-time transcription and translation during class, then post-lecture analysis and feedback.
Reliability and Regulation Matter
As institutions expand agent use, error rates become critical. Each step in a workflow compounds the margin of error.
A suboptimal course suggestion carries different stakes than an incorrect financial aid calculation. Nate Ober, senior ed-tech and AI/machine learning leader at Amazon Web Services, said institutions should ask: "What's the cost of being wrong 10 percent of the time?"
Regulatory compliance poses another barrier. Newer AI vendors may lack experience with education-specific laws like the Family Educational Rights and Privacy Act and the Children's Online Privacy Protection Act. Ryan Lufkin, vice president of global academic strategy at Instructure, said this is especially problematic in the agentic AI space.
"We've got to be very selective in who we partner with," Lufkin said.
Audits and the ability to track each step of an agent's decision-making process will be essential. Nicole Engelbert, vice president of product strategy for student systems at Oracle, cautioned against overstating current capabilities. "Take a side eye on what anyone is saying about what's happening in a pervasive way," she said. "We are in the earliest days."
What Comes Next
The fastest growth in agentic AI for education will likely occur in advising and administrative workflows, where value is clear and risks are manageable.
Burley sees longer-term potential in agents acting as research collaborators and teaching assistants, though those applications remain further out.
Training staff on responsible use will be critical. Ober said: "A large AI investment returns nothing if faculty and staff can't use it confidently."
Institutions should incorporate training into professional development efforts around AI. Success depends not just on the technology, but on the people using it.
Your membership also unlocks:
