How Higher Ed Is Putting AI Agents to Work
At Penn State and DePaul University, AI agents are doing more than automating administrative work. By using AI to connect fragmented data and handle routine tasks, colleges can rethink how they support students in their educational journeys.
This year, Penn State University awarded its millionth degree, a milestone that was deeply personal for David Horton, Penn State CIO and VP for IT. Horton’s father was a first-generation college student and one of the first of those 1 million individuals. It was clear to him that behind that 1 million figure is “the one” — the student at the center of every educational journey. For Horton, the milestone is a reminder that behind every degree there’s an individual with unique needs, and universities must be equipped to meet students where they are.
Still, delivering on this mission of higher education is a complex challenge. Penn State University has over 87,000 students and 37,000 employees. That’s an immense number of students and employees to keep track of and support. So the question becomes: How do you scale AI at an institution of this magnitude?
As AI capabilities have evolved from simple chatbots to action-oriented agents, colleges and universities across the country are increasingly integrating AI into their operational infrastructures. The goal isn’t to simply automate tasks but to streamline processes, provide a better student experience, and give faculty and staff more time for meaningful face-to-face interactions.
During the Education Keynote at Dreamforce 2026, customers shared how they’re implementing agents to transform the student and staff experience. Three key lessons emerged:
1).Data integrity is the foundation for context.
The success of AI agents depends on the strength of the data foundation they operate on. Without accurate, robust data, agents lack the context they need to act appropriately. For higher-education institutions, the problem isn’t a lack of data but rather an overabundance of data across disparate systems. As Horton put it, institutions are “drowning in data but parched for actionable insights.”
For example, Penn State has over 55 systems of record as well as 2.5 million webpages and pieces of digital content. These unstructured, fragmented data systems meant data was essentially floating in siloes, making it impossible for agents to understand how to connect the dots and execute accordingly. With Education Cloud, Penn State was able to bring data together across these various sources and turn it into a unified foundation that agents could reliably act on.
Penn State introduced an enrollment agent powered by Agentforce that saved staff 197 hours during peak enrollment season. As Penn State examined the 79% reduction in cases needing staff support, Horton realized they needed to refocus on the 21% of cases that do need staff support. By focusing on cases that couldn’t be resolved with agents, Penn State was able to think more holistically about the key moments Penn State wanted students to have.
2). Agents unlock time and context for humans to do their jobs better.
Soumitra Ghosh, SVP of Enrollment Management at DePaul University, outlined the university’s launch of a new financial aid agent last fall. The agent was built on both internal data (including previous financial aid data) and external data (including state and federal policies pertaining to financial aid) and could run a 53-point check for every student who inquired about their financial aid. Ghosh said this agent successfully “cut through administrative noise” that humans might struggle to tune out, sifting through layers of policies and procedures to demystify the financial aid process.
DePaul’s financial aid agent has saved over 100 hours in staff time annually and reduced processing times by an average of 90%. These savings opened up more time for faculty and staff to have conversations with students that actually mattered. Instead of spending time going over administrative details, they could focus on actually connecting with students at critical touchpoints. As Bob McCormick, VP of Information Services at DePaul, puts it: “This isn’t about shorter appointments; this is about maximizing that time.”
McCormick noted that insights from agents equipped staff with a clearer, data-informed understanding of each student’s needs before stepping in. DePaul started taking a risk-based approach to assessing student progress, using agents to determine which students were at a higher risk of not graduating. DePaul could lean on agentic tools to resolve issues for low-risk students and proactively provide human support for high-risk students, with a stronger understanding of their needs based on data. As McCormick put it: “Technology is giving context; humans are delivering the message.”
3). Institutions must adapt quickly to keep up.
Higher-education institutions haven’t historically been built to scale technology as quickly as corporations, but they can’t afford to wait.
Horton outlined how Penn State implemented the three most recent Salesforce projects: “The first one we scoped, it took 120 days from launch to go-live. Then we did an add-on piece to it, and that second one took 98 days. And then we decided to bring our bursar into our enrollment service center, and that one took us 80 days. The days of a year and a half for implementation are gone — we do not have time for those kinds of timelines in our industry.”
Additionally, Penn State participates in a Salesforce project, “Quantum Leap,” which allows the university to work directly with Salesforce’s AI professional services teams. Embedding Salesforce teams in Penn State’s AI efforts speeds up experimentation and turns ideas into action faster.
To keep up with the quick pace of AI development, Horton recommends taking complicated problems and breaking them into smaller, more manageable ones, noting that he thinks AI in education is a “permanent disruption.” In other words, institutions have to start somewhere, and they need to start now.
The upshot
At DePaul and Penn State, implementing AI reduced bureaucracy, decreased processing times for students when trying to access benefits like financial aid, and unlocked more time for staff to provide meaningful human-to-human support when it matters most.
As Horton put it, colleges are in the “business of changing lives and lifting entire families up,” a mission he came to understand through his own family’s experience at Penn State.
The promise of agentic higher education, then, isn’t simply doing more with less. It’s using technology to make the human impact of higher education possible at a much greater scale, to improve the lives of students and staff alike.
