How the U of A is Pioneering Agentic AI

```html

One of the challenges faced in higher education IT hasn't been a lack of powerful tools, but the gaps between them. Student-supporting staff members often navigate an array of disconnected platforms to understand a student's circumstances, and then a different system entirely to act on them. Agentic AI offers a new answer to that problem by building a data layer and an interaction layer that can move collections of data to and from source platforms while providing a user interaction or workflow that is platform-agnostic. The University of Arizona is among a growing number of institutions exploring this potential. 

Going Beyond Generative AI 
Most people think of AI as a chatbot: You ask it something, and it writes an answer. That's useful, but it stops there—it can't actually go do anything for you.

The bigger problem isn't a lack of chatbots—it's that getting real work done means moving across systems that don't talk to each other. Updating a record might touch our CRM, then a ticketing tool, an email, and a spreadsheet—each with its own login and its own version of the truth. The work itself is a multi-step, cross-platform slog: gathering the information, reconciling it, acting on it, and following up. Right now, a person must stitch all of that together by hand.

Agentic AI is aimed at that problem. Instead of only answering a question, it can carry out the steps of a task on your behalf—pull up a record, update a status, initiate the next step, and confirm that it happened—moving across systems the way a person would, but without requiring someone to manually shuttle information between them.

In a tool like Risely, you might type an ordinary question—“Who hasn't responded to their advising outreach this month?”—instead of building a report by hand. That's generative AI performing the language understanding. The agentic layer is what finds the answer across systems. Alternatively, the agent might identify a group of students who need follow-up, while generative AI drafts a personalized email for each student using their current records for a person to review and send. Generative AI handles the language. Agentic AI, with a human in the loop, handles the reach and follow-through.

Done right, agentic AI doesn't replace the systems we already use. It sits on top of them, handling the multi-step work of getting something done across platforms so a person doesn't have to perform that stitching manually.

The university's exploration of vendors in this space led to Risely AI, a startup founded in 2025 with a leadership team focused on challenges like these. Risely's architecture centers on a data layer that ingests information from a university's existing systems and uses it to power AI agents that take meaningful action: flagging at-risk students, drafting personalized outreach, generating advising notes, and writing records back to systems of record. The flag itself is straightforward analytics—a score crossing a threshold. What makes it agentic is what happens next: The system turns that flag into a sequence of actions by drafting the outreach, routing it for review, and logging the result without requiring a person to operate each step manually. Risely's goal is to become the operating system for higher education—not a separate platform that staff must log into, but a capability embedded directly within Trellis so staff can work with it the same way they work with any other Trellis app.

The Arizona Online Pilot 
In early 2025, the university launched a focused pilot with Arizona Online, its fully online degree program. The pilot supported student success coaches who managed large, distributed caseloads and had limited visibility across systems. Risely's agentic AI technology was used to connect three core platforms—Brightspace, Trellis CRM, and SIS enrollment data—through a single interface organized around the student success coaches' actual workflow. Trellis CRM is the only system in which the agent both reads and writes data. Brightspace and UAccess are read-only. 

The platform surfaces students most in need of attention based on configured triggers, including declining GPAs, negative early progress reports, and enrollment holds. From a student's briefing page, a coach can review a consolidated history from all connected systems, draft a personalized outreach email informed by that data, and send it directly through Outlook, with the interaction automatically captured in Trellis. During meetings, a voice transcription feature captures conversations in real time and populates structured notes and action items, turning a documentation burden into a timelier feedback report that can be sent to students. 

Early Results and What's Next 
The pilot produced a 32-fold increase in outreach productivity, allowing student success coaches to meaningfully reach more students than previous workflows allowed while increasing student response rates. More consequential data, however, is still being collected. Retention and re-enrollment figures for fall 2026 will provide greater insight into what the pilot ultimately delivered in terms of student outcomes. The university hopes the data will reflect meaningful improvements in students' academic success. 

The Bigger Vision 
Arizona Online is a starting point. Work is underway on Risely projects supporting student service case management for the Office of the Registrar, Office of Scholarships and Financial Aid, and Bursar's Office. The Arizona Online project is also being extended through academic advising pilots in the College of Information, the A-Center, and the College of Social and Behavioral Sciences. During this next phase, agentic AI capabilities like those Risely is building will operate as apps within Trellis. 

Throughout the agentic AI pilot, governance has been a guiding principle, not an afterthought. Before the pilot launched, the university engaged its security office, procurement team, student data steward, and Office of Responsible AI to establish clear guidelines around data ownership, access controls, and vendor obligations. Those agreements were formalized at the contract level—a model the university intends to carry forward as the platform matures. 

The technology is advancing rapidly. The university's pilot has demonstrated that agentic AI is production-ready, practically impactful, and architecturally sound for use by a large public research institution. The opportunity now is to build on that foundation deliberately, responsibly, and at scale to advance the university's mission. This approach can apply to any university administrative process—not only student success coaching, advising, and case management, but also research administration, business offices, and financial services. Anywhere a workflow crosses multiple platforms is an opportunity for agentic AI. Reducing administrative overhead creates more space for high-quality, informed experiences for students, staff, and community partners.