With the proliferation of generative AI like ChatGPT, Gemini, Claude, and others, students now have access to answer giving machines. This deprives students of the opportunity to engage with the material, critically think, and problem solve. AI is here to stay and it is, therefore, our responsibility as educators to harness this powerful tool to help our students actively learn outside the classroom with the help of AI that helps the learning process by personalizing its approach to each student to ensure peak engagement with the material.

We will demonstrate how AI can be customized for teaching and learning using BearcatGPT, UC’s private large language model. For students these AI agents provide on-demand academic support, deeper learning, critical thinking, and higher engagement. For faculty, these agents can be teaching assistants, course navigators, or anything else a faculty member sees as an aide in their classroom.

This workshop will showcase two categories of customized AI agents for:

1. Broad institutional uses like tutoring, test preparation, research, and more
2. Narrow uses like course specific agents that are customized for specific courses and the faculty teaching them.

We will also demonstrate how different AI agents designed for different uses, can collaborate behind the scenes to accomplish tasks that are different in scope but often need to be done together to achieve a goal. For example, when a student shows up for tutoring, and has no foundational knowledge on the concept they need help with, there needs to be a teaching intervention first. To address this, an agentic workflow can be visualized where the AI Tutor agent identifies that the student needs a mini-teaching session and communicates with an AI Teacher agent to teach the student. Once done with the lesson, the teacher hands the student back to the tutor to continue tutoring the student. This workflow adapts to the needs of the student by a carefully designed interplay between both agents to ensure a good mix between mini-teaching sessions and tutoring sessions. This approach leads to a higher level of student engagement and a deeper learning experience. 

We are partnering with faculty across different disciplines like chemistry, physics, law, history, communication, sports administration, healthcare management, Information Technology, and more to build course specific agents. Additionally, we are also developing agents that are not tied to specific subject/content areas for more broad uses like workshops, research, meeting minutes, and more.

Learning Outcomes

Attendees will

1. Understand the difference between generic AI and customized AI. Furthermore, they will understand how customized AI can elevate learning and teaching.
2. Differentiate between broad use agents and course specific agents and when each should be used.
3. Understand design elements such as scope, behavioral guardrails, models, integrations, and more that are responsible for how the AI agent behaves.

The session will conclude with an interactive discussion, aimed at inspiring participants to build their own AI agents within their academic and operational practices.

By the end of this session, attendees will be able to explain how customized AI agents can move generative AI beyond “answer machines” to support active learning and deeper engagement. Participants will learn to differentiate between institution-wide agents (e.g., tutoring, test preparation, research support) and course-specific agents tailored to a particular discipline, instructor, and set of learning objectives. Attendees will also identify key design choices—such as intended role (tutor vs. teacher), scope, guardrails, and workflow integration—that shape an agent’s effectiveness and responsible use.