Loading summary
Transcript47 lines
- [00:03]
A
This session was recorded live at the 2026 ASU GSV summit in San Diego.
- [00:14]
B
Thank you for joining us today. I'm Tim Renick. I'm the executive Director of the National Institute for Student Success and a professor at Georgia State University. Very glad to welcome you to this morning's session. It's been a long conference and it's good to see the enthusiasm still there. We give special credit to our panel today. They're very brave to show up on the last day of a conference on AI to talk about AI. So we're going to have to scramble a bit to find some things that you haven't heard before. We're going to focus a little bit on specifically the perspective from practitioners viewpoints at universities, trying to provide some guidance on where AgentA AI is heading and some of its applications and usage. We've got a varied and talented panel here representing both industry and education and nonprofits. Very quickly, to my direct left is Emily Smith, Principal and VP for Partner Success at College vine. Next to her is Carolina Recci, co founder and co CEO and at ed sites, next is Christina Yancy, VP for Employment and Economic Opportunity at American Institutes for research. Next is J.C. bonilla, chief operating officer and head of AI at element451. And at the end there is Lev Gonick, Chief Information Officer at Arizona State University. So again, a talented group with varied experience, especially in the post secondary space. And so we're going to have a conversation and I'm going to open things up by asking a question to the panelists first to get a level setting of what we mean when we talk about agentic AI. And I think Lev, you had volunteered to maybe do a little bit of the table setting here to just explain a bit what it is and what it isn't.
- [02:10]
A
Thanks Tim, and congratulations to all of you for surviving another ASU gsv. And for those of you for whom it's your first, we hope we'll see you again next year. Tim, the conversation around the Gentec AI is more meaningful this year than last year. And in fact two years ago we even didn't have the vocabulary. I just want that to sort of sink in for a moment. So we are on a journey together to try to find words to describe the very quickly evolving technological capability and what we've done in the first three years of the journey, which is sort of when the world Woke up to AI after 70 years of it being largely inside labs. We've in many ways started with impenetrable use of language. What is Generative AI. And we spent three years ago for a year and believe it or not, of course, all the way to today, folks still wanting to have that conversation about what is generative AI. And folks again, I think have largely said, all right, whatever it was, it gave us a window into a conversation that you could have with a very intelligent, you fill in the blank, a calculator, a stochastic parrot. But almost none of it was like what is the generative part of generative AI? When we started talking about something other than that, we had to find some additional language to talk about it. And that's because as the labs were developing the large language models, the first thing that seemed to break out was this idea that in fact we could have a, an ability to do something that felt magical along with it. But it wasn't going to just be about magic over the long run. There's still a novelty to a lot of that early work. But as we were going to look to go from novelty to normal, we were going to have to find some additional language as the capabilities moved from just being novel language to actually being as it were, leaning into things that would take advantage of the maturing reasoning capability of the language model environment. And that reasoning capability, in my view, is effectively what is different from the first chapter, the generative chapter. And we've decided to call it agentic. The origins of agentic as a language goes back to Tim Berners Lee, for those of you who know the founder of really W3C, the folks who gave us the web browser, his insight was back in the early 90s that the idea that we could actually take away some of the tedium associated with the way we as humans interacted with machines wasn't going to be accomplished 30 years ago. He said that's the aspirational goal. The goal is actually to aspire for a semantic web. This is his language, a semantic web in which agents could take advantage of the abilities of machines to work with machines and thereby liberate humans to higher order function. That was the vision that he called when the agents are able to do that, agents and machines. And so that's the origins of agentic AI in many ways goes back to Tim Berners Lee's framing of it. And what it comes down to here is of course, the ways in which we as humans can help design workflows that leverage advanced reasoning capabilities that allow us to solve both for an efficiency play, which we'll hear about in just a moment in terms of the efficiencies and the things that matter to the university world around recruiting and retaining, as well as the idea of thinking about a new way in which we can evolve with the machines as co workers. And that is sort of the agentic AI journey that I think all of us in our industry are at.
- [06:23]
B
Okay, thank you, Lev. And MIT analysis claims that about 70% of organizations are already in this space, either piloting, launching, or scaling some form of what might fit under this category. So it's a higher level of autonomy for AI and potentially higher production and outcomes can be produced. And so that's where I want to start with this panel and ask them to talk a little bit from your perspective about what the potential is and what the reality is. What kind of outcomes for education can agentic AI produce in your field, in your world? Emily, you want to start?
- [07:04]
C
Sure.
- [07:05]
B
I think you're good.
- [07:06]
C
Am I good? Yeah. Okay.
- [07:07]
D
Sorry.
- [07:08]
C
Loud in my own head. Yeah, I couldn't agree more with Lev. I think what's really interesting is that the models have gotten so much better. A year ago, we were sitting at a place where if you gave AI a really advanced math exam, some of your high school students might have taken this, called the Amy AI models, about a year ago would have scored between 10 and 30%, and now the models are scoring at 90%. When we talk about the meter metric, which is the length of time it takes a human to complete a task and when AI can work uninterrupted to complete this task, a year ago, that was about an hour, and now we're into 8 hours, 12 hours, 14 hours of AI working on your behalf, uninterrupted to do work. So when we think about the opportunity that we see across the operating system of higher education, and this is kind of the thing that I spend a lot of my time on, not only in sort of enrollment, student success, but also the ways in which we as institutions have created a business model that is really, really heavy. I think we have some really interesting opportunities around. Even if we sort of boil it all the way down to, like, financial solvency of institutions and the financial model of colleges and universities, I think if we look at, if you correct for hospital revenue, higher education is at like a $40 billion deficit. And that's not the cost of teaching and learning, and that's not the cost of facilities. This is the cost of administrative load. So over the last 20 years, we have seen things like faculty members haven't really increased per student. Right. We are like teaching with the same efficiency that we have 20, 40 years ago. But the administrators have gone up doubled, in fact, per student administrators have gone up. And yet we still have this student experience that feels pretty crappy. Right? Like, nobody in this room would agree. Like, yeah, student experience right now is awesome. I remember going to college and being like, what does a registrar do? And I went to a prep school. I went, like, I was uniquely prepared to go to college, did not know what I would go to a registrar. Like, the word bursar still, like, drives me nuts, like, what happens in that office. So nobody would agree that we like Slatten. Thanks, Lev.
- [09:19]
E
Right?
- [09:20]
A
It means purse.
- [09:21]
D
Yes.
- [09:23]
C
Now we know the more you know. So I think we can agree that nobody has created this student experience that feels really smooth. And I think the opportunity for AI agents to smooth this out on the student experience side and on the faculty, staff and administrator side is vast. And I don't think anyone on the stage would say, like, down with the humans, like, only AI agents doing this work. I think we would potentially agree around things like, and I'm interested to hear what Christina says here around, like, our job in college and in higher education is to create humans who are uniquely prepared to go out onto earth and be productive in the workforce and to be productive members of, like, Earth.
- [10:03]
E
Yeah. And I would just like to piggyback though on this in terms of thinking about the specific places to insert this now. So in my work, we're thinking about where in this place of the human agentic AI intersection we can realize that benefit. Right now I think it's helpful to think about this in terms of concrete examples. For example, we're currently doing a project with the Gates foundation on their post secondary credit mobility initiative, where you diagnose first the problem, which is that school administrators don't have the time to be able to analyze student transfers in a way that's expedient and smooth for the students, the transfer institutions, and we know hundreds of thousands of students are transferring every single year. This is a big, consistent historical problem. We're finding that if you can have an agentic AI trained to be able to look at equivalency, you can then reposition your human intelligence on the reviewing and acceptances. So I think part of this is just making sure that we discuss this in this way of how we balance this idea that love brought up around the enabling of your human talent in addition to the agentic AI. And I think one of the struggles is that there's just this kind of efficacy cell around the agentic AI as this replacement, and then it struggles to do that and we expect that it's failing.
- [11:33]
D
But it's like that.
- [11:33]
E
It was never intended for that. So I feel like it's helpful. Just as we continue to talk about specific examples, to note that human agentic AI interaction is so critical.
- [11:43]
B
Jc, go ahead.
- [11:44]
F
Hi everybody. JC Element. Couldn't agree more with the flow as it relates to outcomes and the definition of agentic. This synthesis of the very Latin, almost eloquent version of what agentic is. We look at our 400 plus portfolio of partners universities and agentic translates to labor. That's actually what Emily and Christina are talking about. Labor, labor. And of course this starts positioning the framing of the things we can achieve for certain outcomes and oh my gosh, the displacement that comes with work and taking my job and whatnot. So the, the part that I've been seeing that is really effective as we try to push certain outcomes through agents in agentic frameworks, it's that the pattern of data exists in academic institutions. Everyone has data on the student whether it's in a CRM like ours or Excel sheets. Because a lot of our schools still operate that way with Excel sheets, duct tape and prayers. But the data is there. But what they cannot achieve is the ability to action on those signals that come from. From the data that is there right at the pace the expectation that the student wants. Because it always happens somehow. Got it. Apparently I do not know how to hold the microphone. You always learn something the. What was I talking about?
- [13:18]
C
Signals.
- [13:19]
F
Signals. Thank you very much. Just a test. You guys engage. I know. So the ability to action on those signals. It's the difficult thing. And this is where agentic is becoming a reality for student outcomes. Two quick examples that I'm seeing in our practice. It always happens at three in the morning, six hours before the midterm. Right. And also like that, you know, 247 true opportunity to engage the student in their own terms is the reality. That's just basically the low hanging fruit. But from there on we start seeing the potential that. I don't know. First generation freshman student, when she logs into the system at 3am in the morning realizes that she missed the financial aid deadline. Classical example of retention or churn in the students. So now you have these agents that would have truly opportunity to reduce the missed. What happens today to now the action of these signals in the terms that the students want. It's really exciting to see the possibility of these things being enacted and the time from signal to action being reduced. Use cases are endless. We can talk probably all of us hours of the use cases we see into our practice. But that's the generalization. Agents labor and reducing the signal to action time significantly.
- [14:40]
D
Can I add something real quick? Because you said the word signals, which I think is so important in this world of agentic AI, is you got to have the right signals. And I think that's like the foundational layer that these agents act upon is really, really important. And that's something that we're seeing just become more relevant than ever before. One thing that I've found interesting in our work is how do you use agents not just to automate and streamline the workflow of acting on the signals, but actually using agentic AI to capture those signals. And that's something that's pretty new in the space. A lot of times, you know, we work with 300 partners. The signals are pretty consistent of what they're looking at. They're looking at, you know, midterm grades, which are early, but not as early. They're looking at missed payments, registration. But as you said, Emily, the student experience is so much more. And we don't really know at scale in real time how students are feeling, how are they doing with belonging, homesickness, basic needs. And agents can be really powerful with that. If you have best case, and I know very few institutions that have 200 students per advisor, that's a really great ratio. Even then, how are you going to check in with every single student every week? And agents can make that possible and then create the signals because they now can analyze these huge bodies of unstructured data. And. And once you have those signals, pass those to the CRM or to the agent that then will automate not just the data collection, but the action as well. So just wanted to double click on the signals piece because I think we're able to close the signal to action time so much, but how do we make sure that we're chasing the right things? The right inputs, I think, is something that agentic AI can actually help with.
- [16:36]
B
Yeah. So at Georgia State, we've gotten to the point where AI can walk the student through the entire FAFSA completion process. So talking about a couple hours of acting autonomously, we are using it increasingly in the classroom as a supplement for our lack of staffing. Large class sizes in freshman level math and English classes, now in part aided by AI coaching and so forth in these courses and so forth. It is empowering in the short term, but it also raises certain concerns. And Emily, you cited the fact that there are these reports of the flowering of administrators, Although some of those reports are counting administrators very, very broadly to include things like academic advisors and so forth. How do we begin to think about this on our campuses without it becoming a political issue of concern with people, unions, faculty, staff, and increasingly students being highly concerned that every step we take in a positive direction in this space is potential loss of a job for somebody who currently is working at one of our universities or for the students we want to be graduating. I mean, how do we begin to navigate those conversations?
- [18:00]
A
Lev, you want to go at asu, perhaps? Well, we know we're an outlier, but I'll frame it this way. At asu, we're scaling. So for us, the question is, how do we achieve scale? And we know we can't actually do it in the same fashion that we've achieved the scale today. So in the online space, last year we passed 100,000 online learners. President Crow made it clear that by 2030 we will have 200,000 learners. You can't get there from here the same way we got to the first hundred thousand. Again, built into the DNA of ASU is the idea of how do we use every tool that's available to us. It's very clear that agents, as it relates to, again, the recruiting, the retaining, helping students move through the journey in ways very similar to what my colleagues have been sharing here, 24 hour service, when you hit peaks, being able to have the agents wake up, to actually take care of those peak loads without having to staff for peak load, when in fact that's only two weeks a year when you need them, and trying to figure out how to arbitrage the labor challenges that are out there at asu. The answer is there's enormous opportunity here to unleash human agency on the journey that we're on around agentic AI. The problem, again, to go back to my first framing here, is that the language is very, very awkward at the moment. And I'll just ask my colleagues, for those who are interested, to pass along some of the stories that we've been working on at ASU that relate to those twin challenges. One is to figure out how we're going to scale with impact by leveraging the technology. But in the same sentence, talk about the ways in which we are unlocking human agency. And if we don't frame it that way, then the gallows is where we're going to go. To answer your question, Tim, there will be people with pitchforks out there in one way, shape or form across the entire sector. And so again, for the Vendor community who are trying to work with us. The understanding that in fact, what folks are selling can't be efficiency, full stop, because actually that is a significant shortcut to a death knell.
- [20:19]
C
It's leverage.
- [20:21]
A
And we have to be able to be talking about again the human agency. So for those who are interested, I'll just ask Bonnie and Annie just to pass around some of our stories because I won't get to much of them today. But it's that framing of an and both from the beginning, absent that, and we are predisposed in our sector, especially those of us on the stage and most of you in the audience today, to some way or another come back to our significant enthusiasm for and exuberance for the way in which technology can solve. But we've always been in a very important political context for the way in which technologies unfold. And that move from novel to normal, we have to realize, has to be an and both situation. Taking advantage of machines for the things that we now call Agentic AI and unlocking human agency from the 400 universities
- [21:21]
F
that we serve that are unfortunately very far away from the ASU reality, the 1 percentile in use cases and modern deployment of AI. This is not a technology problem. And I want to be very clear with this, that what we're starting to see, and it's very obvious, is that it needs to be addressed as an organizational design problem. Working with agents, deploying agents at scale, you don't solve those issues, as you mentioned, with the tools of the past or what has gotten us here. And academic institutions and our partners are really struggling to see this because the agent shows up and the first thing that it triggers is, oh my gosh, I'm going to lose my job. There's fear and certain blockers are being enacted, so these walls need to be removed. But it's not a technology problem. What is keeping me up at night is that the delta between where the technology is at and where the practice is is significant. Few things that are really, really working with us is two examples that I bring from our practice. What does it mean to work with agents? And introducing agents in the org chart. And then all of a sudden we see community colleges in North Carolina that are saying, hey, how do I onboard an agent? How do I fire an agent? These truly human questions that they may have on how do we develop capacity to get passed to these technologies or design questions or community colleges. Let's stay on that topic. That it is known for them that the admissibility issue is not a problem. One of the dirty words used in community colleges and I hope if you're some of them you will laugh at this, is that if you have a heartbeat, you're in. So it's not an issue of admissibility, it's an issue of retention and persistence of the admissions less office. Think about what that means for the academic institution that has its aspects in how do I get in and whatnot. The admission list office multiple of our community colleges in Texas are already playing with that. It's an organizational design aspect and starting to build capabilities is probably the hardest thing at the moment. Not the technology problem per se.
- [23:32]
B
Okay. And Christina, you're going to say yes?
- [23:34]
E
Yeah, I was going to share some thoughts particularly related from the reflection of thinking about student and learner success specifically so in targeting the use of these agentic tools not just for resolving institutional issues like in terms of understaffed or to address some of the institutional issues, but to be explicit about using it to help address student specified challenges. So I can just give like as one example, we had recently finished a project looking at the challenges of student success and post secondary learners, particularly for women of color. And we were exploring specifically through around 400 interviews and surveys around what the specific retention issues were and trying to resolve some of those to figure out where to help resource answers. We then analyzed 350 institutions websites and found that they lacked totally some of the answers to priority questions. So even if you train an agent to call or pull from the website, if that resource isn't there, it's not going to pull accurate information or prioritize the kind of information related to student support services that specific populations are interested in. So I think to be able to articulate solutions that also are embedded in being able to use these tools to identify problems and then link those to student success and solutions is really important. Bringing that transparency to student success I think more explicitly is going to be key as well.
- [25:22]
B
Yeah, I think that's a good observation. That a lot of the institutions nationally are not in a place where AI can pull from its systems, from its infrastructure. They're siloed. They don't know the answers to the questions as an institution. And so they're at least a generation away from being able to actually leverage what we're talking about here.
- [25:42]
A
And Tim, I think we should just double down on that because the gnarly question is data, data, data integrity, the fidelity of the data. The biggest challenge is when the data that you have in your ecosystem actually contradicts other data that you have in Your system. So it's not that the AI, the agents, are bad news, we're bad news. Yeah.
- [26:00]
E
But the challenge, though, is that the agents, as we all know, have a black box component to it. So the transparency on what those inputs and then those outputs are, are just that much more critical because of the fact that you just don't even know where they're coming up with these answers.
- [26:18]
A
The great expectations is that agents are gonna solve everything. What I'm just trying to say is they're not. We're the challenge. And so we have to figure out how to solve the data problem. And again, Tim and I have been in this industry for a very, very long time, and it's always been about a problem we're gonna solve tomorrow. Hello. If you think agents and, and agentic AI is going to be a path to wherever nirvana, we better solve the data problem together.
- [26:42]
B
But Lev isn't part of the issue. The challenge for maybe some people in the audience, that if you don't have the integrated systems, you don't have the clean data, that it's not just that the agentic AI will not make things better, it may well make things worse. Right. You know, these.
- [26:59]
A
Reputationally, that's a huge issue for those of us who want to champion the ways in which you can leverage technology to achieve these fantastic aspirations, work with partners to make it happen.
- [27:08]
B
It's very advantageous is its ability to scale answers and deliver them efficiently. If it's scaling and delivering the wrong answers, then that's, you know, potentially a disaster for our campuses. I know on some campuses we're working with, at my institute, the biggest block to moving in this direction is legal. You know, it's either legal at the state level or legal offices on the campus that say we, we can't empower you to turn on these systems because we don't trust that it's going to be representing and providing the kind of support that the students would actually benefit from. Go ahead, Emily.
- [27:44]
C
Well, I think you're identifying a war on two fronts. One is content driven and one is data driven. Right. Like where AI agents have to have to Christina's point, like excellent access to excellent data. And it's very usual if you're sitting there going like, holy shit, all of the orphaned web pages on my site and all of the bad information, that's normal, right? That's like table stakes. That's every institution, every time. And then some of the things that Lev is talking about is around sort of the structures of, on average, even small colleges are using between 10 and 30 enterprise data systems right where we have different pieces of data and then a ton of tribal knowledge about what's in those systems and how that's used. And I think there is some advantage of like yeah, I watched an AI agent do what I spent 10 years of my career doing which is go like pick up non relational data, make sense of it, map it and migrate to a new system. And I'm like watching this through like clouded tears, watching unidata like crumble before my eyes. And that's like a thing that can happen now which is cool but there's then like the meaning making around it to like to get to a different point on it I think.
- [28:46]
F
But there's a, there's a shift in yeah, we're not going to get to nirvana. But I do see maybe because I'm a techno optimist, a recruiter, an advisor, a teacher today they're not recruiting, they're not advising, they're not teaching. The way someone spends time on campus today it's on the organizing of how do I get a recruitment event up and so I can come and recruit. 10% recruitment, 80%, 90% organizing. Same thing happens to the teacher and the advisor. So one of the things that perhaps can happen is as agents are taking these low level tasks, the time needed to make data integrity better and these very difficult problems towards nirvana it's attainable. So optimistic view is that I finally see a place for the canonical practice of what is needed in the jobs, teaching, advising, so on and so forth. Now we have a fair chance to do it because right now we're not doing those activities.
- [29:41]
B
Yeah Carolina, we're talking a lot about the efficiencies that come from accomplishing tasks that campuses have long tried to undertake. Part of what you're doing is exploring opportunities to use AI to discover things that campuses have never really discovered before. So why don't you talk a bit about that?
- [30:00]
D
Yeah, I mean I think the reason why that excites me is because you can do it regardless of how your data is set up. So it's if you're using the agent to go talk to the student and uncover the barrier and then it can also solve for it, then that's a new data point that just didn't exist. And I think it's really interesting because it gives us a window into how students are feeling that we've just never had before. We've just never had the ability to check in with every single student on a weekly basis, ask highly Personalized questions about how they're feeling about different areas. We spoke about it last time, but there is really four that we focus on. So wellness, finances, academic perceptions and belonging. To be able to just ask questions around that that are not stuck in a survey or with institutional research and then have that same agent that asks the question, address it, or flag that student for an advisor or a coach is huge because it's kind of. It can operate a little more independently. You don't have to have like a perfect foundational layer. And I think it going back to the fear of what is this going to do to our workforce and what Lev was saying, I think it opens an opportunity for humans to actually there's going to be a premium on the human labor. Southern New Hampshire is a partner of ours and they've done a big campaign that I kind of pitched to other partners now around the advisor role and work of the heart and work of the hand. The advisor is doing work of the heart and then they're doing work of the hand and right now the work of the hand of. Let me register you for a course or answer a question that is taking away from the work of the heart that they probably signed up for to begin with. That's probably what drew them into education. And so if we can have the agent resolve those questions, but then also do the data collection and figure out which students the advisor needs to spend that work of the hard time that makes the advisor so much more efficient because they can just prioritize their conversations in a way that's more high impact.
- [32:07]
B
We're getting down to time here amazingly fast. So what I'd like to do at the very least is have each of you, especially now that we're at the last day of ASU gsv, offer some practical advice to the audience. We know people are in very different places. Very few organizations are where ASU is with regard to this work. Some are just starting and contemplating and so forth. What practical advice would you give to the audience to take back with them to their institution, organization and so forth about a practical step to be taken to be responsible and effective in using agentic AI? Who wants to go? Okay, yes, I'll go first just because
- [32:53]
E
I just to piggyback on top of what Carolina mentioned about this awesome kind of support tool that they've developed and are pushing out. And I think it sounds amazing because it's collecting information heretofore we have never kind of gotten on a real time basis with so many students. But that it also reminds me that one tool is not going to solve the suite of issues because it could lift up, for example, that there's not enough childcare support. And so obviously the tool isn't going to be able to solve that. So we want to think about how your tools are embedded within a broader strategy and that you're thinking of your suite of AI support functions and so the suite of agents, because one is likely to showcase that there's needs for other ones. And then I also think it's important as a second point to think about how we're validating both the inputs into the agents and then what we're doing with our outputs. Because an unmitigated kind of loop, without having humans come in to analyze the success of the strategy. And also think about how we're ensuring that we're using our human time to expand the inputs that are going into the systems will just be super critical to the success. So I feel like those two things are important takeaways.
- [34:16]
C
Thank you for saying two things, because despite my love for odd numbers, I too am going to tell you two things. And I love what you're saying about that sort of validation, like that validation piece. To slow it down, I think I would invite you to realize that it's the third day of a long and overwhelming conference, and you've been among your people the last few days. Like, you've been with us techie weirdos. And then you have to go back home to your institutions and fight the good fight, whatever that looks like for you and your roles. And I will tell you that sort of thing. One that I care about is that institutions do something. I'm watching a lot of, like, freeze, which isn't the right move. In the spirit of, like, we have to prepare students for the world of work and to be effective humans out on Earth, institutions should be doing something. So doing nothing is now, like, not an acceptable answer. To me, the toothpaste is out of the tube on this. You just have to choose where. The second thing that I'll say is that I'm watching institutions. And I've worked with 600 colleges over the last 20 years, and something that I've seen we've sort of dropped in the last year is requirements gathering. Don't forget that you are technologists at heart. Don't forget that you still can gather requirements to solve a problem. What I'm seeing is at the board and president and chancellor level, this directional, like, holy shit, we have to do something about AI. Please, someone aifies something. And then one of the functional business units aifies something and you don't know what you're measuring, you don't know what you're chasing. So I would encourage you to remember that likely your jobs not knowing each and every one of your roles. Your jobs are to remind the folks that you work with on your campuses to do the thing of gathering requirements. Try to solve a problem, set your measure of success, measure it, come back to it and say we made a bet. Did it work or not? Because I think there's just a lot of quick action around AI ifying things that aren't tied to good requirements gathering. Therefore you're going to diagnose the problem wrong and create bad and weird behaviors. So that's number two.
- [36:17]
F
I follow the song of twos or the dance of twos. One thing not to do, don't start this via an rfp, via an action plan or a task force. I mean I spent all my life in academia. I know we love our task forces. Don't do it this way. The requirement gathering component is clerical and the way I would frame it is look for one problem, expensive one. I don't know, international students. I'm losing 10% of my international students because of the immunization form process. Pick something and ask yourself two questions on it. Imagine that we can solve that, signal that problem within 24 hours and imagine that we will address all the students in that segment. That is the inspiration that you need to start using to do requirements gathering. But then the technology comes at the top. Never start with the technology. Do your journey, do the requirements gathering, but pick one problem, an expensive one.
- [37:15]
A
So I would say keep the conversation going among the tribe to pick up on Emily's piece of it. You're all welcome to come. October 20th the 22nd. A little paid announcement here to ASU to keep the conversation going we have a gathering which several folks in the stage have been act called agentic AI and the student experience experience in which this is a conversation as opposed to just answers. I think that's one really piece. The other piece is just to realize that this is the most extraordinary opportunity for you to help lead, lead your organizations, lead your institutions to be the inspiration to unlock again not just the technology smarts that you have, but also the human agency that you can actually inspire people to achieve and understand that creating an agentic workflow within an existing application, the student information system or the registrar system or the bursar system, that's actually not the end of the road. That is the tech ed journey that we've been on for 30 years inside these individual application silos. And yes, there's some efficiencies to be gained, so don't not think about that. But the architecture of the future is actually flipping this so that in fact our commitments that go all the way back to, to Tim Berners Lee, which is about making sure that the data flow to support the mission. And the mission is student success, not the organization success and not the people who live and run these silos. And that at least at ASU we're calling that the shift from Ed Tech to Tech Ed.
- [38:51]
B
Carolina, some advice.
- [38:52]
D
I mean, I think they covered most of the things. I think the biggest thing, I mean I think the bias towards action. I think what you said really resonated to me. I think folks are a little frozen and it's fine to fail and elevating those who at your institution are doing something even when that thing wasn't successful. But having them present, this is what I tried out. It was a massive flop and this is why. And this is what I learned and what I'm going to do differently. I think if anything that will help a little bit with the change management and getting people comfortable and aspiring to be that person that's being highlighted and at least get people a bit more excited about this being something that they can be good at. But also it's totally fine to learn from. So that would be my last.
- [39:37]
B
And I'll take moderators privilege to offer some advice too. Whether you're at a educational institution or in the industry, put some energy into doing research. I was at a session yesterday where MacKenzie was doing a landscape analysis of the actual randomized control trials and so forth that have been run in AI. And it's a very, very small subset of information, I'll say from the perspective of the development of the field of student success over the last 15 years, it's been really important that we've had partnerships between universities and vendors that concentrated on doing independent evaluative research, being able to show what works, being able to answer these questions. And because if we're going to convince our colleagues across higher education to do something differently, we're going to have to speak their language. And their language is evidence and proof and so forth. Right now, as a sector, we have not done that very well in this space. So there's lots of opportunity for you to show what works and show it in a kind of academic and respected fashion. So thank you very much to our very talented panel for their insight and their passion and please join me in giving some appreciation.