Shasta, the University of Houston’s innovative AI chatbot, is now across 11 departmental
websites, marking a major milestone in its campus-wide rollout. Students, faculty,
and staff can ask Shasta questions and receive instant answers 24/7 while browsing
these sites.
As of early October, Shasta had facilitated 32,034 chats across the 11 participating departments, achieving an impressive 82 percent resolution rate — meaning the chatbot successfully answered users’ questions without the need for further assistance.
The Origins of the Shasta Chatbot
Development of the Shasta chatbot began in spring 2024, when University Information Technology (UIT) issued a Request for Proposal (RFP). By fall, UIT began building the chatbot using the Ivy & Ocelot platform from Gravyty.
In October, UIT started integrating Shasta into several student-facing websites, including Financial Aid, Admissions, Student Business Services, and the Office of the University Registrar. These departments were selected for their direct impact on student success.
By February 2025, Shasta was soft launched on these sites. Phase II began in May, expanding the chatbot to the remaining UH departmental websites ahead of the full launch on August 1.
Unlike previous chatbots, Shasta uses generative AI to craft customized responses based on content from departmental websites. Earlier versions could only respond with pre-programmed answers, limiting their flexibility and usefulness.
Ensuring that Shasta provides accurate information is a collaborative effort between UIT and individual departments.
“We needed to make sure, and work with them [departments], to ensure that their websites are up to date,” said Mark Rosanes, AI Solutions Manager with UIT.
Enhancing Shasta’s Performance
Throughout the implementation process, UIT made several improvements to optimize Shasta’s performance. One key insight was the need for dedicated customer service managers to oversee chatbot operations.
“What we found early on is that most of the people assigned to manage the bot were IT people, and we learned quickly that wasn’t the right way to go about it,” said Rosanes.
“Instead, we wanted to include someone who’s more customer service focused, so we created the role of Chatbot Customer Service Manager. And this would be someone working within the department.”
These managers review chatbot conversations to ensure accuracy and relevance. With generative AI, Shasta continuously learns from website updates and user interactions, becoming more effective over time.
“What we’re finding is that the issue of deploying a chatbot really isn’t a technical issue that we’re trying to fix. It’s a customer service issue that we’re trying to fix,” said Rosanes.
Shasta in Action
Since the start of the fall semester, student-facing departments have reported a noticeable drop in customer service calls. The calls they do receive tend to be more targeted, thanks to Shasta resolving many inquiries directly.
In August 2024, departments received 3,417 calls. This year, that number dropped to 2,807. Student Business Services alone saw 4,658 chatbot messages in October.
Kevin Burns, Bursar at Student Business Services, said the department is using the fall semester to evaluate how students interact with Shasta compared to traditional communication methods.
“At the end of the day, we’re trying to serve the students the best way we can,” said Burns. “We will always need people to do things, but if we can get a bot to assist in the job, that’s the beauty of artificial intelligence, to make our job more efficient and expedient.”
Some departments have seen particularly high engagement. Parking & Transportation received 6,505 messages in October, placing it in the top 2 percent of all universities using the Ivy/Ocelot platform. UIT received 9,980 messages, ranking in the top 4 percent among IT bots.
Ask questions, get assistance, and help shape the future of self-service at UH. Shasta is available now — visit your department’s website or explore others to experience the chatbot in action. Learn more about the Shasta chatbot on UIT’s website.