Last year, generative artificial intelligence (GenAI) had an impactful debut at the University of Michigan. Over the summer, the university’s Generative Artificial Intelligence Advisory Committee published a groundbreaking report, offering guidelines and recommendations for how U-M – and any other educational institution – could thoughtfully welcome GenAI to their community.

A few months later, U-M’s Information and Technology Services (ITS), partnering with colleagues across the campus, released a suite of custom generative AI tools which were specifically designed to meet the needs of U-M’s students, faculty, and staff. This made U-M the first university in the world to launch an exclusive AI platform for its community.

With so many creative minds at the university, it makes one wonder: how is U-M actually using these GenAI tools?

There are hundreds of current use cases, ranging from student projects to staff implementations, but some of the most interesting utilizations are coming from U-M faculty.

One particularly successful example is how the Ross School of Business has incorporated U-M’s GenAI tools into the classroom. Last September, Andrew Wu, Assistant Professor of Technology and Operations, and Jun Li, Associate Professor of Technology and Operations at Ross, decided that they wanted to do a comparison test of the popular Chat-GPT against U-M’s new Maizey tool.

Maizey lets users build their own version of a ChatGPT-like service that can be trained on custom datasets, which they can then share with others. Wu and Lin trained a new Maizey instance on the materials for an undergraduate course in operations management for their test case. Following an encouraging test, Wu and Lin created a special Maizey AI tutor for 313 Operations Management, a class with over 500 students.

According to Wu, the Maizey tutor is “specifically trained using your own course materials. With a click of a button, Maizey can enroll in your Canvas site, then absorb the materials there into a structured knowledge base. Then it taps into that knowledge base to answer student questions.” Students can ask whatever questions they want from the tutor in natural language. The tutor can then respond with answers that come straight from the course materials, and it even cites where the answer comes from.

One of the biggest benefits is that the AI tutor is available to students 24 hours a day, so anyone struggling to finish a late-night paper or cram for an exam always has a support tool at their disposal.

Maizey can easily integrate data from Google Drive, Canvas, Dropbox, and other popular tools. This allows faculty members to set up a Maizey for their classes in a matter of minutes. Wu has even produced a series of videos showing how to create a classroom tutor with a minimal amount of work. (All an instructor would need is a few minutes, their course materials, and a valid shortcode.)

The integration of the Maizey AI tutor into Ross’ Operations Management class was a success. The students in the class asked Maizey over 1,000 questions before assignment and exam due dates during fall semester. Wu noted that, based on an average of 150-350 class questions asked per week, assuming (conservatively) that it takes instructors two minutes to answer a single question, this translated to between 5 to 12 instructor hours saved every week. 

Also, according to a self-reported survey, students who said they used Maizey throughout the class noted an improvement in their grade performance. This improvement, as measured by assignment/quiz scores post-Maizey usage, was as high as 5%.

Wu said their team has “now rolled out Maizey to thousands of students in several core classes throughout the university. Our frontline results show that Maizey works well as a capable system in your classroom – one that has a photographic memory of the course content, is smart enough to pick the right content for each question, and, with the GPT engine, can deliver high quality responses that encourage further learning.”

Bob Jones, ITS’ Director of Support Services and Emerging Technology, says this use case is a fantastic preview of what U-M’s AI tools can do for our students and faculty. “This shows how powerful these AI services can be in a classroom setting,” said Jones. “GenAI literacy is going to be vitally important to the future of higher education. We are prepared and excited to help our campus partners to find creative ways to bring these tools into their classrooms.”

The ITS Emerging Technology team is continuing to work with partners across the university to develop new AI tools for the U-M community. One notable example is LSA’s new Maizey AI Advising tool, a 24/7 AI chatbot they call a “smart sidekick for college life.”

Any faculty members who would like to learn more about the capabilities of these tools can visit the ITS AI Services resources page. They can also contact the ITS Emerging Technology team directly with any questions or ideas for potential collaborations via this form.