Customizing AI for Institutional Efficiency

Gerry Bayne: Welcome to EDUCAUSE Exchange, where we focus on a single question or topic from the higher ed IT community and hear advice, anecdotes, best practices and more.

As colleges and universities face tighter budgets, limited hiring and shifting regulatory demands, the need for more efficient practices is touching every part of the institution.

To this end, artificial intelligence is quickly moving from buzzword to backbone, transforming how campuses deliver services, streamline operations and manage their budgets.

Ravi Pendse: It was never our goal to say, how can we use AI to make ourselves more efficient? How can you use the AI to save money? How can we use AI to eliminate jobs? That was that has never been our criteria. But having said that, we have found that in many, many areas it is actually saving people enormous amount of time, potentially dollars and cents.

Gerry Bayne: That's Ravi Pendse, vice president for information technology and chief information officer for the University of Michigan, Ann Arbor. His institution prioritized accessibility, equity and privacy when building their AI tool called Maizey. Yet these value-driven goals produced unexpected efficiency benefits.

Ravi Pendse: AI is the new UI. I want to repeat that again. AI is a new UI.

So, we built an integration between Canvas and Maizey. And now on our campus, if you were a faculty member and you decided you wanted to deploy tutor for your class, you just do it from Canvas and it takes you four minutes to deploy the tutor. Four minutes. That's it.

And the results were quite interesting. What they found is the students who are using Maisy Tutor to study, his students' grade improved on average of 5 to 9%, But more importantly, the faculty members who had deployed these tutors, they were saving between ten to twelve hours per week in office hours.

So instead of a hundred students standing outside of my office, maybe now there are only seven students who really need help, so I can actually spend quality time with them as opposed to next, next, next. (Laughs) And so students who need help, I can actually make that connection, spend time with them, engage with them. Plus, candidly, it gives me more time to also focus on my research.

Vince Kellen: You can kind of put the use of AI into roughly two categories. Category one is just have to do it because it's going to save everybody time. And so every institution or every organization, every company will adopt it and be reasonably similarly, you know, benefited.

Gerry Bayne: Vince Kellen is the former CIO for the University of California, San Diego, when this interview was recorded. He's now the chief information officer for the Texas A&M University System.

Vince Kellen: Then there is: How do we take advantage of our unique content? Our unique data, our unique processes and our own culture to an extent, and infuse that with the AI and now have something that has some strategic advantage, some differentiation. So one of the first things we did is we created what we call a fund manager coach. It's basically an accounting expert that helps our grant people understand how to process grants. Now some institutions generally process grants in roughly a similar way, but the content they use to educate their people who do this work, the way they organize their people, the way they lay them out in the structure are very different from institution to institution. So we grabbed all of that context. Our policies, procedures, training videos. We have communities of practices that meet around these issues. We captured a lot of those notes, and curated that content. That content, I just can't take and plunk down another university, because we use different terms and ways of describing all of this. And that's an example of what I call a very modestly strategic use of AI, or vertical AI is probably a better way to describe it.

Gerry Bayne: The suite of AI-powered assistants developed at UC San Diego called TritonGPT gave them more than just an operational advantage, it saved them money.

Vince Kellen: Because it's our own hardware, and we, you know, when we include all our costs, electricity, everything, we're still running it at a fraction of the cost of the procurement of a service from a cloud provider. Because it's on premise and it's ours and it's scaled for what we need.

And we don't have investors who've plowed billions of dollars into me who need me to charge my customers a high amount.

Speaker 1: The University of Michigan is also seeing big efficiency gains in procurement thanks to AI.

Ravi Pendse: Michigan has a budget of close to, what, $16 billion or so in our total budget. And so we buy a lot of stuff. So there's internal billing and invoicing that goes on. So partnering with them, our teams worked together and we were able to deploy a Maizy, essentially, tool that does all of that invoicing that this external vendor tool used to do. The other tool used to actually cost hundreds of thousands of dollars. The Maizey tool costs $62 a year. The old tool used to require my three full-time employees because it was 50% accurate. The new tool is 97% accurate. So the full-time employees now are able to spend that time in much better ways supporting other customers and supporting other needs of the organization.

Gerry Bayne: As institutions across the globe use AI to pursue efficiency gains, many are concerned about the environmental impact of building data centers that train and run artificial intelligence models.

Jenay Robert: This question of efficiency has multiple interpretations. And one of the things that I think is particularly challenging for higher education is to think about the environmental impacts of these tools. So if we're increasing efficiency in some ways, but then are we using a tool that is environmentally inefficient?

Gerry Bayne: Jenay Roberts is a senior researcher here at EDUCAUSE.

Jenay Robert: You know, are we trading one type of problem for another type of problem? And so, I just want people to keep in mind as they're thinking about impacting efficiency with AI tools, that that means a lot of different things. And that's another thing that makes it so challenging to do — you're not just considering, are we saving money? Are we saving time in FTEs?

We're also thinking about, are we negatively impacting the environment?

Are we widening any digital equity gaps? Are these models trained in ways that are not aligned with our ethical standards and value systems? And so it becomes a very challenging question, not just of money and time, but of some of these other bigger questions.

Gerry Bayne: The University of Michigan has stated that they have been diligent in making sure that Maisie, their generative AI platform, has the lowest carbon footprint possible, while also providing the best possible tools to their community.

For more information on AI implementations in higher education, visit the EDUCAUSE AI resource page at educause.edu/AIresources.

I'm Gerry Bayne for EDUCAUSE. Thanks for listening.