Compendium of AI Applications in Higher Education

AI Application

Create a general campus information assistant for staff, faculty, and student support

Rationale

Streamlines access to AI tools in a mobile app for around-the-clock answers to common queries about campus or specific courses.

Case Study

In 2025, the University of Michigan (U-M) launched Go Blue, a mobile AI companion that can answer university-specific questions and integrates with U-M’s course-specific AI tutors. Users can access a wide array of information—including dining menus, dining hall hours of operation, student organization updates, shuttle bus information, student support resources, U-M sports, campus news sources, directory information, local weather, campus events, and course materials (where applicable).

Status

Deployed (as of September 2025)

Implementation Details

Go Blue’s GPT is powered by OpenAI’s GPT-4 offering. Information and Technology Services staff maintain a Go Blue App Beta Testing Team to gather feedback on new features and updated information before general releases.

Impact

Medium

Implementation Effort

High

Resource Cost

$$$$

AI Application

Build a platform that extends the capabilities of generative AI models to users

Rationale

Enables any campus user to experiment with generative AI tools, reducing inequitable access to “premium” generative AI tools and ensuring that tool security, privacy, and accessibility align with institutional standards.

Case Study

PhoenixAI is The University of Chicago’s (UChicago) custom generative AI chat service offered to faculty, staff, and students. PhoenixAI provides all UChicago community members with free access to OpenAI’s GPT-4.1 and o3 LLMs via a user interface that, unlike most commercial offerings, meets A11y standards and, especially, the University of Chicago Digital Accessibility Standards.

Additionally, by hosting the tool on the University’s protected environment in Microsoft Azure, their user data is more protected from third-party commercial use than if they had purchased enterprise licenses from OpenAI. Users can decide which model (GPT-4.1 or o3) to use at any point via a drop-down menu, use internet search to inform model output, analyze data, write software code, customize the chat tool to their needs, and more.

Another benefit is that within PhoenixAI, users can create, share, edit, and delete custom “AI Assistants,” which use Retrieval Augmented Generation (RAG) with user-provided files to enhance model responses. Users commonly use this capability to create course-specific tutors or chatbots for use within a research group.

Status

Deployed (as of September 2025)

Implementation Details

UChicago staff partnered with a vendor, Royal Cyber, to develop and launch Phoenix AI in September 2024. They built PhoenixAI in the Microsoft Azure environment, and the first release used OpenAI’s GPT-4o model.

By working with a vendor instead of exclusively using in-house staff to develop the tool, UChicago was able to launch the first version of PhoenixAI within six weeks of signing the contract. A cross-functional group of staff across the institution works on the regular update releases to fix bugs and test new functionality and/or features in collaboration with Royal Cyber. The UChicago team includes project managers, business analysts, designers, software developers, accessibility experts, and information security staff.

As of August 2025, PhoenixAI allows users to toggle between GPT-4.1 and GPT-o3. UChicago pays for access to OpenAI’s models via a token-based pricing scheme. To cap their expenses, they provide students, faculty, and staff with model-dependent query quotas. Users are notified when they reach their daily limit:

• GPT-o3 model: 50 queries per day limit for faculty and 10 queries per day for all other users
• GPT-4.1 model: 50 prompts per day for all users

Staff at UChicago are working to update PhoenixAI to include a new security role, Personal Health Information (PHI) Restricted, which would enable university researchers to share sensitive data with PhoenixAI while complying with HIPAA data privacy standards.

Impact

Medium

Implementation Effort

Medium

Resource Cost

$$

AI Application

Create institution-specific AI agents for campus constituents

Rationale

Facilitates the safe and effective use of AI tools by providing community members with purpose-built AI agents optimized for university tasks in a protected environment

Case Study

PatriotAI is George Mason University’s (GMU) enterprise AI platform that grants students, faculty, and staff access to six institution-specific AI agents in a secure environment:

• PatriotChat: serves as a conversational assistant for general George Mason information.
• DocuMate: summarizes, derives key takeaways from, and compares scholarly documents.
• PatriotPal: accurately answers student questions about university-specific administrative procedures, academic policies, and services.
• NourishNet: connects students experiencing food insecurity to food access programs, offering a stigma-free experience.
• CourseMate: helps students understand course content, analyze reading materials, and prepare for exams.
• SyllaBright: supports course design, helping faculty create and refine course material (e.g., syllabi, planning assessments)

Status

Deployed (as of August 2025)

Implementation Details

PatriotAI is an instance of nebulaONE, Cloudforce’s multi-modal Generative AI platform built on Azure. It is a product of GMU’s partnership with Microsoft Education and Cloudforce. GMU describes this as only the “first lineup” of AI tools available on the PatriotAI platform. Community members and leaders on GMU’s AI Visioning Taskforce gave significant input into the development of the first six agents currently featured on PatriotAI. Agents within PatriotAI use Brave Search API for general internet and targeted site search queries.

Currently, GMU students, faculty, and staff can sign in with their university Microsoft account to access the beta version of the platform. Leaders plan to debut the platform at scale in fall 2025.

Impact

High

Implementation Effort

Medium

Resource Cost

$$

AI Application

Provide instructors with course material design support

Rationale

Saves instructor time by streamlining course and instructional materials development directly within the LMS (e.g., course structure, test questions, assessment tasks)

Case Study

The University of Leeds was one of the first institutions to enable the AI Design Assistant feature in the LMS Blackboard Learn Ultra when it first became available in September 2023. Based on existing content in the course, the AI Design Assistant provides instructors with AI-generated recommendations for adding new modules, generating test questions, developing authentic assessment tasks (based on Bloom’s taxonomy), creating grading rubrics, and adding images to course content when prompted. For most content types, users can input topic guidance, desired difficulty level, and complexity, etc.

In the first five months of its release, 95% of instructors responding to in-app feedback surveys indicated that the AI Design Assistant was saving them time, and 92% would recommend it to their peers.

Status

Deployed (as of September 2025)

Implementation Details

Blackboard Learn Ultra’s AI Design Assistant is powered by the Azure OpenAI Service. Anthology (Blackboard Learn’s parent company) and Microsoft’s product development staff convened 19 clients from their AI Advisory Council, including University of Leeds, to develop and test the functionality of the AI Design Assistant before its release in 2023.

Impact

Medium

Implementation Effort

Low

Resource Cost

$

AI Application

Provide tailored instructional feedback to educators

Rationale

Supports consistency and accuracy in supervisor feedback and evaluation processes for students, specifically teacher candidates.

Case Study

Sibme’s AI Insights analyzes uploaded lesson recordings (video or audio), instructional materials, and student work to develop personalized, evidence-driven feedback on educators’ teaching practice, including timestamped video observations. Coaching feedback analyzes and emphasizes key instructional metrics, such as Teacher Talk Time, Student Talk Time, and Question Distribution.

Leaders of Martin Luther College’s graduate and continuing education teacher training programs implemented Sibme’s AI Insights to provide student teachers with instructional feedback (e.g., where lesson plans miss key objectives, where students were not engaged). Sibme’s AI insights and summaries reduce the time instructional coaches spend on reviewing student instructional videos, freeing up more time for higher-level coaching and mentorship.

Status

Deployed (as of September 2025)

Implementation Details

Sibme is an instructional coaching software company founded by a former Texas teacher in 2013. They provide technological solutions for coaching, monitoring, and assessing the teaching efficacy of K-12 educators, which higher education leaders can use to enrich their teacher training programs. Developers at Sibme partnered with engineers at AWS and integrated AWS AI technologies (specifically, Amazon Rekognition and Amazon Comprehend) to begin offering AI-powered features in 2024. Sibme’s data strategy and policies align with FERPA, COPPA, and GDPR privacy regulations.

Impact

Medium

Implementation Effort

Low

Resource Cost

$$

AI Application

Supply 24/7 personalized real-time tutoring

Rationale

Provides students with one-on-one academic support at times when previously not possible, enabling instructors to spend more class time on student relationships, hands-on practice, and learning.

Case Study

The Stephen M. Ross School of Business at the University of Michigan launched a virtual Teaching Assistant (TA) pilot program in 2025 to facilitate self-paced learning. Instructors can customize the virtual TA to their specific courses and teaching style, so that it can explain complex course concepts to students and work through difficult problems without directly giving students answers. The virtual TA is powered by Google’s Gemini models.

Status

Deployed (as of May 2025)

Implementation Details

U-M (along with several other institutions) has piloted the virtual TAs in various business courses (e.g., financial technology, operations strategy, and operations analytics and statistics) at U-M. Early pilot results show increased student engagement and deeper instructor understanding of student learning patterns. The virtual TA delivers real-time data on student usage and frequently asked questions, enabling instructors to simplify and improve teaching materials. Leaders from Ross plan to incorporate insights from the pilot program into a study co-led with Northwestern University’s Kellogg School of Management, involving 9,000 students, 72 courses, and 26 schools.

Impact

Medium

Implementation Effort

Medium

Resource Cost

$$$

AI Application

Solicit small gifts autonomously

Rationale

Expands donor reach by using AI agents for autonomous frontline fundraising.

Case Study

Texas State University (TXST) enlisted Givzey’s Version2.ai to deploy autonomous fundraising agents referred to as Virtual Engagement Officers (VEOs). Without any staff intervention, VEOs qualify donors, build relationships, and solicit and close gifts.

Within just one week, TXST’s VEO, Emma, reached 925 donors, had more than five touchpoints with 70% of the assigned portfolio, and raised $4,000 from 78 gifts.

Overall, Emma independently completes the work of over 20 gift officers by sending personalized messages to 2,000 donors in its portfolio, analyzing donor behavior for tailored engagement, and responding to donor inquiries in real-time.

Status

Deployed (as of September 2025)

Implementation Details

Version2.ai enabled TXST’s advancement team to launch VEOs in just three guided weeks with a total time investment of around eight to 10 hours. Staff onboard their autonomous fundraiser with institutional information and the advancement team’s processes and goals.

Impact

High

Implementation Effort

Low

Resource Cost

$$$

AI Application

Guide drafting of procurement statements of work (SOWs)

Rationale

Enables procurement specialists to draft SOWs quickly and consistently, addressing common pain points such as a lack of commodity knowledge and unclear milestones.

Case Study

Procurement specialists at California Polytechnic State University, San Luis Obispo (Cal Poly) use ScopeBuilder: an AI-assisted tool that streamlines the development of statements of work (SOW) for contracts, bids, and RFPs. ScopeBuilder guides university procurement specialists through the SOW drafting process step-by-step (e.g., suggesting milestones, drafting work scopes in real-time) based on user input, established best practices, and relevant past contracts. ScopeBuilder ensures compliance with institutional standards, offers support for buyers with limited familiarity in a commodity area, and incorporates templates and preferred language from exemplary historical documents.

Status

Deployed (as of September 2025)

Implementation Details

Cal Poly’s Chief Procurement Officer led the development of ScopeBuilder in partnership with a team of students and Amazon Web Services (AWS) partners from the university’s Digital Transformation Hub (DxHub). The DxHub is an AWS-powered innovation hub that provides students with experience solving real-world problems in collaboration with AWS.

ScopeBuilder is primarily powered by Anthropic’s Claude Sonnet 3.5 model via Amazon Bedrock. All functionality and services exist within the AWS ecosystem to ensure confidentiality of university procurement processes. Student developers built the front-end using React (hosted on Amazon S3 cloud object storage). ScopeBuilder also uses Amazon CloudFront for content delivery, AWS Certificate Manager for encryption, and AWS Lambda for computing functions.

Impact

Medium

Implementation Effort

Medium

Resource Cost

$$