Unpacking AI at Scale with CSU and OpenAI

Sophie White: I am thrilled for you to check out our episode of EDUCAUSE Shop Talk today. This one was really fascinating. In it we talked to implementers from the California State University system and from open AI about what is actually the largest deployment of ChatGPT as we found out during the show in the world because the Cal State system signed a partnership with OpenAI to implement ChatGPT across all twenty-three of their campuses. So we really dive in today about why Cal State decided to make this decision, what it had to do with closing the digital divide, making sure that all of their campuses with various different resources had access to this, why it was important from a privacy perspective for them to implement a closed system, the really thoughtful change management approach that they took to implementing this large deployment across their institutions. And also through working with faculty, staff and students to make sure that they had agency in how the rollout went, how they used data at a system level to evaluate what was working, what wasn't.

Working with the ChatGPT rollout and really making sure that they took time to play and have fun with this implementation to make sure that folks understood the value of it. So it's a really great conversation. I think that there are lessons that state systems that larger systems can take from it, but also that smaller institutions who maybe don't have the scope of resources as Cal State but are still interested in doing some kind of partnership or creative work with AI can learn from as well. So check it out. Let us know what you think.

***

Sophie White: Hello everyone and welcome to EDUCAUSE Shop Talk. I'm Sophie White. I'm a content marketing and program manager with EDUCAUSE and I'm one of the hosts for today's discussion.

Jenay Robert: My name is Jenay Robert. I'm a senior researcher at EDUCAUSE and I'll be your other host today.

Sophie White: Great. So we're really excited to have a special conversation today. So we really started this podcast to talk about big issues in higher education and technology and today we were talking about a very big deployment related to higher ed and AI. So we're thrilled to have a lot of the players involved in this conversation here today to talk about what that looks like. So we'll be talking about the California State University system and their large deployment of generative AI tools in higher education. So we'll talk about the ongoing system-wide deployment of ChatGPT, which is being done across twenty-three campuses and supporting more than 460,000 students and over 50,000 faculty and staff. So big deployment. We have several guests with us here today representing OpenAI and the Cal State system from both the system-wide offices and one of the institutions using the ChatGPT deployment.

So I'll introduce everyone and then we'll jump into it. First we have Siya Raj Purohit and if you want to just wave for the folks watching, that would be great. An education leader, author, and investor. Siya is with OpenAI as the education GTM leader. She is on the education team at OpenAI where she helps manage the deployment of ChatGPT Edu across higher education and K through 12. So thanks Siya for being with us. Next up we have Dr. Ed Clark, who is CIO for the CSU Chancellor's office. Ed is an award-winning leader, skilled at leveraging technology and relationships to develop innovative approaches for strategic outcomes. Ed's currently the CIO for the California State System, the largest university system in the US, which I just mentioned, comprised of 23 campuses. It's a lot over 50,000 staff and more than 470,000 students. Ed's role is to serve as the strategic partner for the chancellor's executive team and collaborate with university campus presidents and their chief information officers to catalyze strategic and innovative administrative and academic technology initiatives to advance the mission of CSU. Ed currently also serves on the board of directors for 1EdTech and CENIC, the Corporation for Education Networks Initiatives in California, which are organizations at the forefront of educational technology and networking. Thanks Ed for being with us.

Ed Clark: Thank you.

Sophie White: Next we have Dr. Felix Zuñiga, campus engagement partner on the ChatGPT implementation team. Felix is a campus engagement partner at the Cal State University and works in the office of the Chancellor. He focuses on engagement for the CSU Tech community. Felix is passionate about family, education, leadership, mentorship, and student success. He's a past winner of the EDUCAUSE DEI leadership award, past chair of the DEI advisory committee, current faculty of the EDUCAUSE Management Institute, and also an alumni of the Next Leaders Fellowship. You do a lot for EDUCAUSE too. So thank you Felix, and thanks for being with us today. And last but not least, we have Dr. James Frazee, CIO at San Diego State University. Dr. Frazee is Vice President for Information Technology and Chief Information Officer at San Diego State. A collaborative leader with over 25 years of experience at SDSU, he's championed transformative AI initiatives, including the development and deployment of the largest AI student survey ever conducted in higher education. Under his direction. SDSU is partnering with the Cal State University campuses and institutions worldwide to adapt the survey for local use informing the development or refinement of their AI strategies. You're doing a lot with AI outside of this partnership too, so thank you James for being with us.

James Frazee: My pleasure.

Sophie White: Alright, so we have an all-star group here today. Let's dive into it. So I'm curious, how did this idea come about? Who kind of started, let's create this massive deployment of ChatGPT at Cal State and what did that look like?

Ed Clark: I can try to tackle that at least from the CSU side and it'd be interesting to hear the OpenAI side of that too. We had several faculty resolutions in the CSU on AI in 2019 and again in 2023 that there was an urgent need to explore artificial intelligence and how it could be leveraged in higher ed. So we formed this generative AI committee that worked over the semester, met many times formed with subcommittees and around teaching and learning around enhanced productivity around security and privacy. And one of the main recommendations was we need to ensure that we have inclusive and equitable access to AI tools in the CSU. And why that's a big deal is across the system. We were seeing each university start to investigate their own contracts with Microsoft, with Google, with OpenAI, but some of our smaller campuses couldn't afford to do that, so they were just going to be left behind.

And building on that there were also, we knew there were thousands, thousands of people and I was one of 'em that had their own subscriptions to each of these tools. But that data wasn't protected in a way that we could get that we wanted to have. We wanted that data protected for our community. And so we were looking for a partner. We were going around looking for partners actually at EDUCAUSE. I started having this conversation with lots of different partners and we were so fortunate to find a willing partner in OpenAI to actually think about can we somehow reach an arrangement that would make this achievable for the CSU because we're worried about a digital divide. Hey, you know what, if the prices remain the same and the deals are just the way they are, a large portion of our community was going to get left behind. They wouldn't be able to afford it. And yet we think this is a technology that five, ten years from now, if you'd say you're not going to adopt it or use it, it'd be similar to saying you're not going to be using the internet. So it's something we did not want our community left behind and I'd like to have my colleagues here add to that as much as they want to.

Siya Raj Purohit: I can share the OpenAI perspective and honestly what I love about this partnership is just what it does for workforce development and economic mobility for all Californians. The CSU system coming on board is huge for this and enabling all these different campuses to get access was very powerful. On the OpenAI side, we launched ChatGPT Edu in the summer of 2024. So as you all know, ChatGPT launched in November, 2022. We saw a huge uptick from professors and students and recognize the need to enable a connected campus experience so professors can create things for their students in a safe way. Staff and administrators can use it with university data in a safe and secure way. And so ChatGPT Edu is designed to be very secure for our campus. None of the information in that workspace is used to train OpenAI models and admins have a lot of controls on what information gets shared within the workspace and outside of it. So then professors and researchers feel much more comfortable uploading their research work or class curriculum knowing it won't leave the workspace. So we launched that in the summer of 2024. We have worked with over a hundred institutions of all sizes and K-12 school districts and others as well. But CSU is very unique because when it was announced, it was the largest deployment of ChatGPT anywhere in the world, including our enterprise customers. So really impactful to be able to work with this amazing team.

Jenay Robert: That's so exciting to hear that, I didn't realize this was the largest deployment in the whole world and that's just another way that CSU is so special. I always appreciate the work that you all are doing. And I also wanted to really highlight the importance of what you talk about, Ed, with different size campuses and differently resourced campuses having different access to these tools. And this is something that we're seeing in our research at EDUCAUSE and our landscape study that was published earlier this year that we, it's bearing out in the data that we're seeing that institutions that have different resource levels, though they're all just about equally interested in adopting these technologies for all the reasons we've talked about, workforce development and educational equity, so many things that they can support. Institutions simply don't all have the resources to do it. So I think this system wide approach is something that a lot of other systems can learn from as well as consortia and so forth.

Felix Zuñiga: I'd like to just add to that Jenay, that equity is really at the center of what we were doing. We needed to make sure that students had access right from Humboldt in the north to San Diego and the south. And a lot of times we do initiatives or we say you have to do this, but then we leave it up to the campus to figure out how to pay for it. And so the partnership that Ed talked about creating really allowed for us to pay for a tool that everybody doesn't have equal access to all faculty, staff and students and really took that kind of barrier off of the table. There's a lot of other challenges that we will probably get into talking about today, but at least access to the tool was a primary thing that we were looking at. And James, maybe you could talk about what that looked like at the campus level, right? Or what did you feel when you heard that this came down right from the chancellor's office?

James Frazee: Yeah, we were thrilled in terms of the ability to democratize access to the high quality tools. Many of our students, we looked at the data, in fact, the number of students, faculty, and staff who signed up for the free ChatGPT and paid versions of chat GPT with their SDSU email, which by the way, we say don't do that, use your personal email address. But because they did that, we had visibility into who was paying for it, who was not paying for it. And we saw a dramatic difference in that kind of haves and have nots in terms of the pro version of the tool. And so this was an opportunity for us to democratize access to the more robust, more accurate pro versions of the tool. And we saw a remarkable uptake. We rolled this out right before summer break and we have, as of this morning 19,338 active users, call it 20,000.

And it breaks down to be about 50 percent of our employees, both faculty and staff, and 44 percent of our students. And that was with little to no advertising right on the eve of final exams and the run up to a lot of high stakes assessment. So this semester, as we are about to launch the fall term here, we're working with our associated student government leaders to share a video that they're producing to talk about our generative AI privacy statement. One of the things that students are concerned about is being surveilled. The same can be said with faculty and staff. That is absolutely not happening here. And so we went to great lengths to put together a privacy statement with our university senate, our university legal counsel to make it exceedingly clear to people that we are not monitoring either what you're prompting or the output of that.

And we talk about how to use these tools responsibly. And we can say more about that as we go on through a number of different mechanisms, including microcredentials that are part of our first year experience. So all of our first time freshmen go through a one unit seminar and part of that is a requirement that they go through an academic application of AI micro-credential, which really emphasizes how to use these tools responsibly. So this is another opportunity for us to really plug that micro-credential and the importance of that, especially for our transfer students who don't get to go through our first year experience. And so it has been really important, this partnership has created space for conversations with the community about what we know about how people learn and the importance of active learning and how if the instructional design is done well, you can avoid some of the concerns that people have about academic misconduct. So that's been really exciting as an outcome

Jenay Robert: Since you brought up the, oh, go ahead Ed, please.

Ed Clark: Oh, sorry. I'm so sorry. I figured that you might want to know system-wide, that system-wide, similar timing for across the system. We now have 104,000 activated accounts and that was in a very short time right before summer. So we do expect a big surge this fall.

Jenay Robert: I wanted to ask, since you brought up concerns related to academic misconduct, this is something I'm really interested in in terms of the big differences we see among the higher ed communities. Some people are saying generative AI tools are completely going to revolutionize the way we teach and the way the world operates. And then we still have a fairly large portion of people who are, we can say resistant to generative AI tools, still banning their use in the class. And I'm curious, especially being such a big community, how did you navigate that wide range of opinion?

James Frazee: Yeah, I can give a couple examples from San Diego State and I'm curious to hear from our other panelists a couple ways that we went about that. One of the things that I think is the most important job of the CIO is to provide lots of examples, especially when emerging technologies are taking the world by storm, such as they are with generative AI. And so we have launched an AI and action video series where we feature faculty talking about how they're using generative AI, why they're using generative AI and what were the outcomes both for them and for their learners. And I think it's really important to have those disciplinary examples for other faculty to learn from. And it's also an opportunity for us to celebrate and recognize those faculty who are kind of the early adopters and the pioneers and not only to recognize them but then to give their colleagues an opportunity to reflect on the work of their peers.

So that's one of the things that we've done to address that. The other thing we're doing is we're incentivizing faculty engagement. Maybe our chancellor's office colleagues can say more about the challenge grant program that they rolled out, which effectively was seed money that we're building on here at San Diego State. This year we're launching a new initiative called AI across the curriculum where we're paying faculty, which is an important sign of respect for the labor costs involved here when they have to reimagine their assessment strategy. We're paying a faculty member from every major at San Diego State to reimagine an activity or a graded assignment using generative AI. And as part of the deliverable for that, they will record one of these AI in action videos. So we'll have an example from every single major on campus of how people are using AI. And again, that does a couple things. A) They get support so they're not just left alone to figure this out, but they're getting support, concierge support, I would say from our instructional designers and from our AI faculty fellows. And it also is an opportunity for us to make sure that AI is woven throughout the curriculum so you don't have to be a computer science major or a computer engineering major to get exposure to how to use these tools responsibly.

Ed Clark: And I think the question about how to navigate that tension between the supporters and the detractors of our strategy, I'll just say that we are obviously still navigating that just in case anyone thinks that we've got it solved. As a system, it's an ongoing conversation and I think it really gets to, in building support for this strategy, whenever we talked about the fact that our students are going to be expected to know how to use these tools by employers, it really does change the conversation. And so I think it's hard to deny that because we've actually had the conversation with our employers on our AI workforce acceleration board. So it's not something that can be argued. Yes, a big chunk of our employers are expecting these skills to be understood by our students when they come out of college. And so that helped us kind of navigate through that.

And to James's point about our grant program, I wish my colleague Leslie were here, but we had an AI educational innovations grant program, and we did a call for proposals between, I think it was April 1st to April 30th, so just a month. And we had over four hundred submissions with hundreds of faculty coming together. These weren't one-off submissions, these were submissions that were supposed to be work with colleagues to figure out how you could transform a program, a course, that kind of thing. And so having hundreds may have been thousands of faculty together, I don't know the actual numbers, but showing that level of interest was also very powerful for the rest of the community to see, hey, my colleagues are really interested in this and they're trying to do this. Maybe I need to rethink that as well. But it's still an ongoing conversation. And Felix, do you have more to add? You may have better data than I do.

Felix Zuñiga: Yeah, I was going to say 750 faculty were represented in those over four hundred proposals, which total three million dollars. I know, like you said, I wish Leslie was here because she talked about when they put that out there, they were sure how many people were going to apply and to receive four hundred proposals. They weren't quite ready to look at all of that, right? So it's an amazing thing that so many people were looking to collaborate and do this work and getting ready to dive in and do this. So I think it's just going to be an even stronger second round, right, Ed, we can find some more funding out there. If anybody watches this and has money they want to help, we're always accepting that kind of seed money for these proposals.

Siya Raj Purohit: And then the opening eye front, basically if you have realized that professors are the real influencers on the campus, once they adopt AI, thirty to three hundred students in their classroom engage with AI too, and they're the ones who can influence other professors within their disciplines to start adopting the tools thinking about changing curriculum. So some of the interesting things we've seen is that professors are rethinking assignments in the way that midterms, for example, used to have A and B, they would rotate every year so students wouldn't cheat and have the paper from the year before, but now they can create twenty-four or thirty variations of that exam and have it very custom to what the students care about. So you're learning in a much more personalized way than ever before because AI is helping them with that experience. These kind of changes are super interesting.

OpenAI is also trying to help promote professors who are doing interesting things, showcase their work to others across disciplines. They have a large OpenAI Academy and an OpenAI forum where professors talk about what is the future of math in the o1 reasoning era, how do we teach digital marketing with AI? So super fascinating conversations designed for all professors. And the second thing we're seeing a rise of as a teaching tool is AI reflection forms or basically AI disclosure form system called them. But basically this is a form attached with assignments that professors share with their students, and it's like a five minute exercise where students share how they used AI, what worked, what didn't work, where AI helped them and where it didn't. And this provides unprecedented insights into how students are thinking about their assignments. We never had this before because you never knew how the student actually came up with the work that they submitted, but these AI reflection forms are giving professors this interesting viewpoint, but also letting them guide their students on how to use AI more effectively. So very powerful. I think we're going to see a lot more of that this fall semester.

James Frazee: Yeah, we've had a number of faculty create their own custom GPTs with content that they're curating, which is freeing up, especially there's one faculty member who recently spoke to our council of vice presidents about their experience with this, and this is a person who teaches two 500-student sections of psychology 101 every fall and every spring. So the person has about a thousand students, and they talked about how helpful this was for students who needed that just-in-time on-demand support. They launched it this summer, they'll be rolling it out, scaling it up here in the fall. But they talked about how this was basically a TA 24/7 TA for these students and especially for the neurodivergent students and students who had other challenges they were facing, this was seen as a real equalizer for them. So that was something we were excited to hear about.

We actually had an OpenAI hosted in an enablement day here at SDSU back in April to showcase some use cases. And those demos for our faculty in particular, I think really went a long way in terms of getting them excited about what might be possible. And I think faculty in particular need evidence to grasp the magnitude of what's changing before they're ready to change themselves and their behavior. So that's where the AI survey that we've rolled out has been so important because we can provide data from our campus, here's what our students are saying, here's what our students' expectations are for us as a university, here is what the employers are expecting of us as a university. And so that really does go a long way in terms of seeing faculty start to think about the various possibilities.

Siya Raj Purohit: One interesting thing about what James mentioned about these 24/7 TAs custom GPTs is that most students are using them between 12 AM and 2 AM when a human teacher is not available. So it's fascinating that it just expands the kind of support they're getting from these professors.

Ed Clark: Building on what Siya just said, so many of our students are working students too. So I teach at Cal State Fullerton, I teach information systems. I'm teaching students starting at seven, a class that goes from seven to 9:30 at night, but they've already worked. A lot of 'em have been working all day. They come to this course and then if they don't understand something at 11 when they get home or whatever, it's a hard time to get support. And it's such an important point for so many of our student bodies across the country.

Jenay Robert: I love that you brought that up because I was a working student as an undergraduate, and as Siya was talking, I was just thinking back to those overnight study sessions at Denny's where you meet up with a couple of friends and overnight's the only time you have to cram for that exam the next day, and oh my gosh, if we had had a tutor sitting there with us, that would've made all the difference.

Sophie White: Yeah, I love the focus on equity and just thinking about this larger time that we find ourselves in and that we're really proving the value of higher education. There are a lot of headwinds that we're facing right now. I think it's so important to be thinking ahead and how can we support non-traditional students? How can we create equity among the system? So thank you for sharing that. I have a question, which is whatever you're comfortable with sharing. You mentioned getting a lot of proposals that being more than you expected, maybe an unexpected challenge in this implementation. Love talking about all the wins, but are there other challenges that you did not expect that you ran into as you were implementing this? And do you have lessons for other institutions who might be able to learn from you?

Ed Clark: Absolutely. And I'm going to ask Felix or James to support me. I'm just the implementation challenges. This was new to OpenAI, just like it was new to the CSU, so trying to create these workspaces for every university had its own separate space, but then we wanted to be able to, what if we built a really great tool that all students should have access to? How would we share that tool across multiple universities? That ended up being one of the hurdles that we were trying to figure out. That was a challenge. I know that as we moved folks from their paid versions to their university versions, there was a, could you migrate your past chats into it? These are very, at the technical level, there was, and then what happens when you leave? If you leave the CSU, will you be able to take your chats back out of it? And there's interesting things like that. James and Felix, can you remember more than I'm bringing up? I'm just thinking at the top level here of what I,

Felix Zuñiga: Yeah, I mean, James, we all teed up for you. I think change management and communication were the biggest things, right? Because I think, and you might've said it, James, when you thought you've communicated enough, it's not enough, right? I mean, please expand on that if you can.

James Frazee: Yeah, I like to say when you're tired of communicating, you're just getting started. And that's absolutely been the case here. And part of the reason we've seen such a tremendous response rate to that AI survey I mentioned earlier is because of the shared governance that has been really the foundation of this work with our associated students. So I mentioned them producing a video about that AI privacy statement. Having the message come from the students is really important. We like to say our faculty development is by faculty for faculty. So we've got AI faculty fellows who are respected faculty members who are often leading that charge. That has been really essential in terms of some of our successes, in terms of some of the things that we might not have expected that was revealed through the data gathering. Some really dramatic differences between groups. And so for instance, male students were more comfortable with AI than our female students, dramatically more comfortable, which was interesting.

International students were dramatically more comfortable than AI than domestic students, graduate students versus undergraduate students. Our San Diego students versus our students out in the Imperial Valley, which is we have a campus 110 miles east of here in a very rural agricultural area. And so some really dramatic differences between groups. And again, that was data that we could use to scaffold conversations with the community of, okay, now that we know this, we have to act on that. How are we going to act on that information in a way that is going to help these students to balance that and find that equity? The other thing that was really troubling is one of the questions we asked was about number of smart devices. So if you own the laptop and a desktop and a smartphone and an Xbox, but basically the more devices you had, the more positive you perceived generative AI as influencing your learning experience. The more comfortable you were with the tools, the more tools you were using, the more likely you were to pay for those tools. Not a big surprise, but if you use number of devices as a proxy for privilege, the more privileged students were at a major advantage over our students who had less resources. And so that was evidence of a growing digital divide. And so that was, again, more information for us and really points to the importance of this partnership.

Felix Zuñiga: To add to that, I would say when you have an implementation of this size around 500,000 people, there's a lot of voices and a lot of opinion. And so you may think, I've talked to groups or we've moved this forward, and until you're talking to every faculty member, it's not enough unless their voice is heard or any staff member who has concerns or any student who doesn't want to be lumped in with everybody else, these are all valid concerns that we need to be listening to and hearing. And we need those voices at the table because that's most important thing is that we need to have the conversation because if you're not a part of the conversation, you're being left out and others are being making decisions for you. And so just to talk about the technical side, we work really closely with Siya and team. We meet on a weekly basis. We have open lines of communication back and forth. And there's times when we're like, well, we want to be able to do this and we can't because it's a new product, it's bleeding edge and higher ed is not used to being at the bleeding edge. A lot of times a times we're slower to adopt technology. So we see that as, okay, not only do we have a chance to use the latest and greatest, but we get to help shape what this looks like for the rest of higher education. And that's been really valuable in working with the team because I mean, Siya, I hope that you could talk to what we're bringing forward. You're seeing or hearing from other educational institutions is just at a different scope and level that we're kind of dealing with these things right now.

Siya Raj Purohit: Yeah. What CSU is doing a really good job of is trying to bring everyone along in this process. So different levels of enablement being offered to different to staff members, faculty, admins like IT people, students has been really helpful. So basically training exactly where you're at. I feel everyone actually starts the AI journey in the same way by improving emails. You drop in an email you don't like ask for feedback. That's kind of the small win everyone starts with. But it's been amazing to build on that and showcase how do you get from there to building custom GPTs to building strategic documents, new startups out of just using ChatGPT Edu. And I think CSU is doing an interesting job at training at every level right now.

Jenay Robert: I have to say, I think my ChatGPT journey didn't start with emails, but I think it started with turning requests to take the garbage out into fun things like poetry for my husband. So slightly different use case there, but hey, marital communication, it's important.

James Frazee: Love that.

Sophie White: I think mine started with recipes. I have all this random stuff in my kitchen, what can I make with it? This is great. And I'm curious, so thinking about the technology folks that we work with at EDUCAUSE, do you have any lessons from the rollout of it that you want to share? So did you start with just a small group of users for testing and then expand from there? Or what does that look like that you can share? I know it's still early in the process, but where did it start and where is it going in terms of the rollout and implementation?

Ed Clark: Well, in our, because we're a system, we made sure that every campus could decide how they were going to roll it out, when they were going to roll it out. And certainly at the chancellor's office we're a much smaller crowd. We said, okay, we are the test case. Let's just tackle that and roll it out here and see how it goes. The opt-in was very fast. And so as you could see, we had a hundred thousand activated accounts in just a couple of months at maximum. And I think to your point, this fall, as we see that rollout accelerate even faster, we might run into problems we haven't thought of yet. But I would say in terms of lessons learned from the rollout, I think Felix hit on it that sort of, and James too, the communication, the change management communicate over and over again. We were trying to figure out, we are working together to generate infographics, things that we could post everywhere to show people what was going on and how and when to sign up.

And then having lots of, actually, to your point about using it for fun, having trainings where we could say, I've seen some of these very dry trainings where it's like, okay, here's how to do a really good prompt and all this. But really bringing the fun into it I think was an important aspect of this. We just had this AI in July event where we had these prompt parties and people were working on things like, Hey, generate song lyrics or generate the best picture you can possibly, those kinds of things to match this theme. And I think that's a key portion as well. What do y'all think?

James Frazee: I think the data gathering is so important in having evidence that you can use to scaffold those conversations goes so far, one of the things that we heard from our students is that they're getting mixed messages. In one class, they were being encouraged to use generative AI, in another class, it was completely forbidden, but often they didn't know until they had already used a particular tool and they would find out after the fact. So that actually created space for a conversation with our university senate to put together a special group made up of university senators that created some guidance for the use of generative AI for assessments and deliverables. And that was really important because that ended up in our curriculum guide. So other faculty have some examples that they can build upon or swallow whole if they're so inclined that make it really clear for this particular assignment, it's encouraged on this assignment, it's restricted on this assignment, it's allowed.

So I think it's all about having really clear guidance for the community. The community wants guardrails, they need heuristics, and they need to be part of defining those. We as the administration can't impose that on anyone. They have to craft that themselves. We just need to create the conditions. On the staff side, one of the things that I think is super important is encouraging people to tinker and experiment and have a way for them to do that openly so that it is being encouraged. I mean, we're all facing a dramatic budget reduction. There's a huge structural deficit here in the state of California. So as people are retiring or leaving through attrition, we're in a situation in many institutions where there's a hiring freeze, so we're not able to backfill those positions. So we want to encourage people to work smarter if some of these tools can help with that tier one workload, that frees us up for the more interesting tier two, tier three work that often only humans accomplish today. So that's also important.

Felix Zuñiga: I would add on the technical side, it's exciting that OpenAI is always releasing something new. I mean, last week we got the study mode and GPT-5, and so it's hard when you're trying to draw a line in the sand to teach and roll these tools out and get people prepared, but you're coming out with so many great things, it's a moving target. So that's a challenge. And so luckily, I would say OpenAI has been very, very, just able to be flexible with how rigid sometimes higher ed could be, and not saying, well, you're not going to get access to the latest and greatest because we don't have a new contract or anything like that. No, it's always been, you'll have access to whatever is the latest. But again, that moving target makes it hard for people who may have challenge with change. And so it's like, wait a minute, you just told me this and now it's this, or the instructions say something different. It's like, well, it's new, it's changing. You're on the edge. So it's part of that just enablement of change and learning and constant new information. We talk about AI hitting everywhere all at once in all of our lives. And so people are dealing with it at home with their kids now with work and with school. And so it can be very overwhelming. And I think it's a matter of how do we take that in, be able to listen to people, and help them move along as well.

Jenay Robert: Can we go back to the mention of AI in July? I had the absolute honor, pleasure, joy of being invited to one of these sessions and oh my gosh, it was so fun, including AI generated music that I need to have on my Spotify list. But I remember at the time actually thinking we need to figure out how to get a Shop Talk episode just about AI in July, and maybe we'll do it at some point, but Felix maybe, or I mean anyone, but I'm guessing Felix, if you want to talk a little bit more about that specific event series, I think it's just such an interesting model that other institutions could steal.

Felix Zuñiga: Yeah, no, thank you for that. And we'll gladly share the model with anybody. Essentially, we had our tech conference that we've been doing in person and in 2020 we pivoted to virtual. So as we were trying to figure out, are we going to get back together in person or not, the state says, Hey, you can't travel. So it's like, okay, we'll stay virtual, more people can attend. We had about a thousand people come online before in a traditional all day virtual, but it's very, very tiring with all day Zoom kind of things. And so last year we pivoted and said, alright, AI is like all the talk right now. Let's just do AI in July and we'll do a series of things. And so we ended up doing a series of keynotes, like Ed talked about prompt parties where AI challenges essentially we do campus showcases, we feature what's going on in San Diego State.

We had these technical deep dives for our nerdy tech types who want to learn more about specific tools with the vendor partners. And so it was kind of like something for everyone. And when we really got into this year, we threw everything against the wall to see what would stick and everything stuck. So we had something planned for every day of the month. And so it was a lot, it was overwhelming, but at the end, we had several thousand people come in and they're still going back and watching the sessions. And Jenay, we heard a lot of great things about the EDUCAUSE presentation that you did. People want to learn more about that data, how they can get access to it. Remind me to give you the feedback that we got from your session. We could do that for you as well. But it was a matter of just using the tools that we had, knowing how to engage an audience that we built over time and then doing things that they want to hear about. They want to hear from keynote speakers around, how do we manage humans in this crazy time of change? They want to hear from visionary leaders about the system, what our campuses are doing, like San Diego State leading the way, trailblazing with all the different projects that they're doing. So it's just a matter of getting out there, sharing the winds and helping support each other by collaborating.

James Frazee: Yeah, I would say that event, and I definitely want to hear what you have to say about that Siya, is that teachers, instructors, generally speaking are increasingly on board, but struggling with how to help students use generative AI in ways that don't lead to learning loss or dulling of their critical thinking skills. And so through events like AI in July, you get to hear from faculty who found creative ways to use these tools to get the students diving more deeply into the content or interacting with one another in the content in ways that were never before possible through things like role playing and other more active, again, we're back to active learning. Any new change that gets us an opportunity to talk about what we know about how people learn, I'm all about it. And especially for faculty who may not have gotten a PhD in education who don't understand some of that, this is now creating an opportunity for some of those conversations. So kudos to the chancellor's office for AI in July. It's been a huge hit.

Siya Raj Purohit: One of the activities that OpenAI participated in AI in July, it was a GPT icon, so basically enabling people to build GPTs that help streamline different workflows. So we showcased three of the most successful GPTs across the CSU system and the owners came up, talked about how they built it, how they iterated on it, and then we asked everyone to spend about an hour building new GPTs. We got hundreds of interesting submissions. And it was fascinating to see how there's so much admin workflow that could be automated with GPT to free them up to do the things that they care about more in their jobs. So it was really interesting to see those submissions come in.

Sophie White: I'm curious, can you all talk a little bit about how you're planning to use chat GPT on the administrative side? We've talked about faculty and staff and learning a lot, but what does that look like, especially as we're looking at workforce challenges? And our last Shop Talk actually was about putting people first and partly about how we can do less with less or optimize how we are supporting the institution's mission through our workforce. So can you talk about the administrative side of this implementation?

Ed Clark: Yeah, I'll certainly start. I think that across the system to Siya's point, we already have our campuses going out there and doing all sorts of interesting things from an administrative perspective. We had, for example, I think one of our campuses adopted some AI tool where their recruiters, there are 24/7 recruiters. They can answer questions for students that are interested in their campus at any hour of the day, that kind of thing. There are also really interesting use cases that are important right now. For example, financial aid fraud is occurring across the United States right now. And the way we would do it, the old way to do it is like, would pattern match? Okay, this kind of email, if it has this kind of subject or it did this kind of thing, maybe that's how we identify this and we're actually exploring building a bot to go out there, and I won't name the campus.

We looked at one campus and found a thousand fraudulent applications to that particular campus. This is a really important way that we could use AI to do some of this administrative work that also will protect our campuses and our students and their financial aid and the productivity wins. Certainly many of our developers and AI in July, they talked about this, they're seeing the development process go a lot faster. They like havoc, do coding go after the fact and fix the code. Those are some of the things that I've heard and I'm sure James and Felix will want to bring out more and Siya.

James Frazee: Yeah, here in the California State University, the IT profession has a new set of job classifications that our human resources organizations are rolling out across the system. And so we've built a custom GPT to help managers with writing position descriptions that fit within that new job classification framework. So for instance, now we have a job classification for instructional designer. In the past, if you were an instructional designer, the closest fit and it wasn't really a good fit, was something called an information technology consultant. And so we kind of shoehorned people into particular job classification because there wasn't a good fit. Now that we've got this more robust set of job classifications, managers are still trying to figure out where all of the pieces fit. So that's a really good administrative use case. Of course, that's the first iteration. We're still going back with our human resources professionals to have that human in the loop to trust but verify that that actually is a sufficient job description, let's say for a position as their role is evolving in the university administration. So that's one example.

Felix Zuñiga: I think that's one of the interesting things that we're seeing.

Jenay Robert: Offer a few points that I've seen kind of thread throughout this conversation. And I think really it all comes down to this cultural approach to what you did. You didn't just grab a piece of technology, slap it on a shelf and tell everybody this is the technology you're using. I'm really hearing this coordinated top down, bottom up approach to transforming the culture at your institutions. And that to me is really inspiring. I'm hearing that you took a data informed approach, which as a researcher, of course my heart goes pitter-patter. I'm hearing that you really allowed your faculty, your institutions, departments, colleges to act autonomously to the extent that they could with this rollout. And I think that's such an important piece of how we operate in higher education. We really value that faculty autonomy, disciplinary autonomy, and I'm hearing that come through, engaging faculty as influencers. I love Siya that you said that I want to see all faculty go out and become TikTok stars now. Please, that would be so fun.

And speaking of fun, I think you were all able to make this really fun along the way. And what a challenge to do that when everything is changing constantly and people feel thrown off and people feel like the very essence of what education is being thrown off, out of balance, or they're having to rethink everything they know about teaching and learning and work and everything in between. And you've really been able to make it fun. I think perhaps by communicating when you're tired of communicating, you have just started, right, James? So I mean, those are just some of the key points that I noticed and what in your story, and I wanted to reiterate them mostly because I think our audience might be thinking, we may not have a giant rollout, but maybe we want to try pilots. Maybe we want to try working with just one department. It's going to look different for everybody. And I hope that everyone listening can really take something from this story in just this approach, the mindset, the collaborative approach that you all took. It's really, it's educational. I think for that reason,

Sophie White: I really love the respectful approach to change management too. James, just even mentioning we're paying faculty for their time to learn this new skill. I think it's so essential as we think about all of the changes that we're seeing in the world. We have heard feedback about, oh, these faculty don't want to do something, but let's treat them as people and talk to them and make sure that they understand that we value their time too. So I just appreciate that approach and the fact that you all are iterating and learning and understanding more as you go. I think there's so much we could say about this conversation, but I think we have to wrap up to support everyone's time. But thank you again to the Cal State folks. Thank you Siya, for representing the OpenAI perspective and thanks for kind of sharing the messy during implementation process that you're going through with all of us. We can circle back in a year maybe and see what else you've been up to, but appreciate your time and all the lessons that you taught us today.