The Higher Ed Marketer’s Guide to ChatGPT and Generative AI is a special 4-part podcast series brought to you by Enrollify and Element451 and hosted by Ardis Kadiu, Founder and CEO of Element451 and, Zach Busekrus from Enrollify. 

Over the next four weeks, we’re taking a deep dive into the past, present, and future of Artificial Intelligence's role in higher education marketing and student recruitment. 

The Enrollify Podcast

In Episode 4, Ardis and Zach are joined by Element’s Chief Technology Officer, Petar Djordjevic for a conversation on how Element is using AI to build one of the industry’s most powerful and user-friendly CRMs on the market. Topics include:

How can ChatGPT be used to help you garner deeper insights into your data?

What is semantic search, and why does it matter?

How is Element451 helping their clients/partners better understand how to leverage the power of AI to ascertain deeper, more helpful insights for higher ed admissions professionals and marketers?

How AI can provide better recommendations (6:15)

Generating personalized student communication at scale (9:20)

Closing the gap for understaffed schools with AI (20:00)

Using an AI-powered CRM for human-to-human interactions (28:30)

The Higher Ed Marketer’s Guide to ChatGPT and Generative AI is podcast series brought to you by Enrollify and Element451.

In Episode 1, you got a crash course on ChatGPT and why higher ed marketers and enrollment managers should care about this revolutionary tool.

In Episode 2, you joined Ardis and Zach for a live brainstorm on how marketers and admissions professionals can use ChatGPT to generate innovative campaign ideas and increase operational efficiency.

In Episode 3, Ardis and I were joined by JC Bonilla, Element board member and Chief Data Officer at Vayner Media, for a conversation on the history of generative AI and how the broader advertising space is using AI to promote their products and services.

And finally,  in Episode 4, Zach and Ardis are joined by Element’s Chief Technology Officer, Petar Djordjevic, for a conversation on how Element is using AI to build one of the industry’s most powerful and user-friendly CRMs on the market. 

Well, folks, welcome back to Episode Four in this four part podcast series that Enrollify and Element have teamed up together on As a quick refresher, if you're just joining us for this episode, we're doing a deep dive into AI and ChatGPT in particular, and how it impacts and will continue to impact higher education specifically in the context of higher education, marketing and enrollment management. So this is a powerful like jam packed series. Last episode, we had JC Bonilla, on, who's doing some incredible work at Vayner Media. And who also sits on the board of Element, the episode before that Ardis and I actually do some screen sharing where we go into ChatGBT itself and queue up several prompts and that said, this is a really dynamic series. If you haven't listened to the first three episodes, scroll on down to the show notes below. And we'll have all of those episodes linked there. But this episode, again, kind of wraps our series here. And what we wanted to do and why we wanted to bring Petar into this conversation is we want to talk a little bit more practically and specifically about how ChatGPT and its underlying mechanisms will be used in the context of higher education CRMs, right in the tools that you will use every day to attract, engage, enroll and, and retain your students. So I'm super excited to have you on and again, just super thankful for you making the time to have this conversation with Ardis and I.

Thank you, Zach. Thank you for this opportunity. I like to present what we're building and how we're thinking about AI and I think there's some cool stuff in there.

I couldn't agree more. And I thought it would be nice and a nice way to just kind of kickstart this conversation Petar, if you could just share a little bit about like, what, how is AI already kind of used in the context of most CRMs? Like already, like aI think sounds, it can sound like this very, like abstract, you know, the robots taking over the world kind of thing. But the reality is, most of us interact with AI every single day already with the tools that we use. So could you just give us a quick overview of some of the things that we're probably all used to doing in our CRMs that AI is, is helping facilitate those interactions?

Sure, sure. So, one good analogy that I heard when I was researching a while back is, don't think like AI. If you like incorporating it, it will instantly change your flow, I don't think it will, like make you immediately smarter. Think that like you just got like 1000 hands or like 1000 people working for you and trying to help you to automate certain tasks. So if you think about that, CRMs before us because there's like multiple generations, right? Yeah, as time goes on, the innovation is, especially in the AI space in the last couple of years, it's growing really rapidly. But before you most, most commonly, you will see AI being used in cases of, for example, fraud detection, that was a really common scenario. Detecting duplicates, right, especially in the CRM space, right, because your CRMs are mostly you're dumping your data from different sources, and there might be a lot of duplicates so that it was really good at detecting that, I mean, humans will be able to do that as well. But you'll be able to encode that kind of knowledge. And that was one more use case that's quite interesting as well, was predictive analytics, right, trying to predict what's next based on certain actions is that that that users or students or depending on what, what, what's your actor and the most common that everybody knows off right recommendation systems? Do you see how the Netflix you have in your HBO max or different streaming platforms? So that's what you've most probably seen and use the daily Amazon right shopping cart? What to buy next based on what you did before so that it's quite, it's not something new, right? It's been here for a long time. Yeah. But now there's a lot of hype because this latest generation of innovation that has had Ben makes it really easy to use really powerful machine learning models to power your day-to-day. And then I think you will see that a lot of companies now will start incorporating it. And it will be maybe, I think it will be I have feeling simpler to add. So you can expect tools and products to have this incorporated more easily. Yeah, well going forward.

Yeah, yeah. I don't know. Actually, a question that I have is like, marketing automation, right? Is that term that we've all used for years? Right. And this idea of like, if then logic and like we talked about, like the decision tree last episode, right? Is that sort of like also just a very practical like, it's there's not a human that's going in and sending that confirmation email, right? There's workflow logic that's built that after somebody comes in they RSVP to your own house, they're gonna get a triggered notification saying, hey, Pattaya, your you know, your we've saved you a seat right to our open house. So that is also right, like a form of AI at play. Correct? Or is that technically something else,

At the end of the day, like we, you know, we can call something machine learning AI, we can call it automation, but the outcome and the kind of the goal for the outcome is very similar. We want to automate something, we want to have this helps do things that we traditionally do manually, and when you think about it, is that the machines are really good at detecting patterns. So yes, the automation part can be done in a more simplistic way. However, when you think about AI is its ability to look at a massive amount of data and in the CRM space, or specifically in the tools they're using every day, it's looking at all the activity and the actions that you're taking, and trying to predict what are the patterns, so it can give you like better said, better recommendations, or kind of give you a different view of a particular email or a particular kind of piece of content, depending on your prior history. And that's the things that we have been focusing on the past. And Element traditionally has been around activity information, or behavioral, behavioral AI or behavioral machine learning or data analytics, like the behavior is super important when you deal with CRMs. Because you're, you're capturing the, you know, you're capturing things like addresses, and so on, so forth. But it's the behavior that really makes a difference in how people interact with your brand, and how they interact with you as a school or as a program that is going to say, Hey, you're more likely to open my next email or even more likely to actually receive my call, and you're a better fit than somebody who has not interacted. Yeah. So that's the simple notion of kind of where we've been over the past five years.

So let me throw out this scenario. And you both tell me if this is a helpful way to think about the distinction between marketing automation, as we might know it today, and what when we're talking about AI and its applications in the context of CRM, that might mean, if not today, like in the near future. So let's say right, a prospective student, they go and we'll just stick with this event marketing analogy. They go and they RSVP to an open house event, right? Marketing automation, as we know it right. There'll be one trigger communication, it might include one or two, you know, variable tokens, right in the context of the email, it'll, it'll pull in Petar's name, and also his like program of interest, right in the confirmation email. But basically, there's going to be the same email that everybody who signs up for this particular event is going to get. Whereas as AI continues to be infused into these tools, what we could see is we could see a confirmation email, everyone's signing up for the same event, but the confirmation email that folks get could be wildly different in terms of semantics and in terms of tone and style, based off of what the machine has learned about how Ardis interacts with content versus how pet interacts with content. Is that right, guys?

That's, that's exactly. And one of the things that you can think about that is, you know, 100, hands helping you. Yeah, I keep coming back to that analogy. Well, think about it this way. If I were to respond, you know, individually to that student or to that email. If I were to write that email, I would learn a little bit more about Zach. And kind of know, it's so quick as Zach is in this location. He's interested in this program. And then I would write an email that kind of puts all those pieces together to convey that, hey, this event is going to be very important, and thank you for coming and we appreciate you traveling this much because you're at this distance. So all of those things that make it really personal, well, guess what? The machine is going to be able to generate that at scale, rather than, you know, having the more templatized.

Yeah, and, and when you mentioned like because you know, with a student, and you might know his interests, you can pull extra resources that you have, summarize them and give it that extra, like do to help convert, right? Hey, you know, read this, this might be important before you come to an event. And this can be personalized, depending on the student that are really, really large scale. And that's, and that's that part, right? Because how can you achieve that? With just decision trees, you will still get to a certain limit the number of cases that you can support. Because it does not scale humanly possible to maintain that. Otherwise, you will need to hire 1000 hands.

Yeah, yeah. Do you all think that there's going to be some sort of like, interim step like I'm thinking, right? I know that, some tools do this with like sequences or like email, and automation flows in your CRM, where it's, it's an attempt to be a little bit more personalized. It's basically a template. But then before you hit send, right, it'll say, hey, Zach, we're gonna send this email to Pat are, you know, do you want to customize anything real quickly before you know? Or do you want to just enroll him in this sequence, and then, you know, hit go, and he'll get this automated set of communications, right? So basically, it's a templated, like three-email series. But you have the ability to very quickly throw in one line of personalization if you want before you enroll your contact in this workflow, right? So that's pretty standard, a lot of CRMs have stuff like this, do you all think that there will be missed the step where CRMs will basically like, suggest, hey, we're, you know, Zach, here are three versions of an email that we could send to a pet or right now as to confirm his like registration for this event? Like, will there be some more interaction with folks, at least initially, because I would imagine, like, if I have oodles of data, or like, that's great, and I can pull that from, but if I only have a little bit of data, it's gonna be hard to sort of like, differentiate in a meaningful way, especially if pet RS never sent me an email with like, a bunch of emojis. Right? He might be an emoji kind of guy. But if I haven't, if, if there's been limited correspondence, right, I don't know that I'm gonna get that I don't know if there's gonna be enough data for these tools to kind of like pull from so what I guess what, what is like the interim steps to getting to a point where someone you know, 50 people RSVP to an event and they get 50 different email responses, like what do we think is gonna happen first?

So thinking about this, one thing that you mentioned, is interim steps, and I'll first go to the end. And then we'll work way, way, way back. When you talk about generating the emails, what's really cool feature that will be possible. And this all again depends on the platform, or how the platform is decoded. But these things will be able to generate entire workflows for you, but you won't need to even have a human intervention to create these steps. So that's, that's the future, right? That's like, hey, you know, the student, you know, his context, and you get the workflow specific to that student. Because it's it. It's not just generating text, it can generate text that can be interpreted into actions, and you can, using those actions, create resources, which are like your workflow steps or emails or so that's like the vision, right? Because in the end, you will just have data and you will have algorithms that can, you know, do something to convert that student or optimize for something, right. And they can measure success and feed that they have a feedback loop and get better into integrated content. So you're talking about intermediate steps, what if it does not have enough context? Right? Yeah, have enough information. Good thing is that these models? So there's, there's a caveat, these models have a problem, where currently, there's something called hallucination is and you saw it probably saw that in ChatGPT. It can generate information that's not accurate. It tries to just figure things out from the entire internet, radio, they force you to speed internet and he tries to generate some texts. Yeah. But you can say, hey, don't do that. Because when you're instructing it to do something, sending that prompt, that's what they call it, like prompting the model. You can say, hey, don't, don't go outside of this context of this data. If you don't have enough information, don't try to figure things out, just you know, say, I don't know. And you won't be able to get that as a result as the output of that little black box. It can tell you, I don't have enough information. Action, which can then be a signal that you can code into and be like, Okay, we need to prompt the user to interact, give it more data, and decide what to do. So there will be scenarios where it has enough data. Yeah. And you just leave it do its work or you intervene. Of course, there needs to be a lot of testing and through this, you get, you know, you see that, do you trust this process? Or you can just blindfold it and let it run. But I think initially, of course, while training it, you will intervene. But in the end, it will just work. And it can probably set up your whole system. At least that's promise.

Yeah, go ahead Ardis. And then I have a follow-up to that.

I would just want to say that this is the interim step. This is kind of what we're working on right now. Right. So we have a large language model that can generate stuff for you. But you have to be really good at telling it exactly what to do. Yeah. So you don't get unexpected results. And that's going to be the majority, it's going to be this fine-tuning, and how you built the prompts. And that's a lot of the work that's happening in products right now are around taking that technology and putting the additional layer on top of it, that is going to get you the output and the result. And there's a lot of really cool, from a technical perspective, really cool and challenging tasks that's happening with chaining prompts, the way the prompts are done, the feedback loops that are happening, but you need to have a system that is able to handle all that stuff and able to inject additional context, depending on user actions, like chatbots, for example, it's like, yes, you can answer a question, but it's going to give you garbage, if you don't, you know, if it doesn't, you know, it's going to generate something for you. But it doesn't mean that it's true. So that's where the systems are in place. And that's where the on the product side, we have to be good at figuring out what the context is who the user is, how to grab additional context. And so that's the immediate step right now that a lot of the work is going towards, it's like in the product itself.

Yeah, yeah. No, that's super, super interesting. And like, when you were saying, Petar, when you were talking about sort of like the also sort of the constraints almost, that you could give these tools of, like, hey, only look at like this data, right? To help it make, you know, facilitate your answer, whatever. What's super interesting about that, and I hadn't thought about this before, but like, we're where we're probably also going is like, hey, let's only look at prospects around these particular like programs, right? Like, let's say you're working with a graduate program, right? And you don't you don't necessarily want to analyze sort of every graduate prospect and every graduate project prospective student, because depending on the program that they might be interested in these individuals these personas, right, they might act very differently, right. But it's like, but if you only want to pull from, Hey, what is all, you know, what, what history do we have of people that are interested in, you know, an MBA program, right, and they're from these particular, like, you know, regions of the world, and let's only pull from that historical context, right? To inform this workflow that we want to create, versus let's pull from all the data that we have across every program that, you know, within the School of Business, to write these workflows. And that is super interesting because maybe that's also sort of this, this medium step of before getting to we were talking about this, in our last episode, we were talking about sort of like a segment of one, right, like a next sort of being like the ultimate goal of, of how to best leverage these tools is how do you deliver the best possible experience down to the individual right, to promote your product or your service, and that's maybe where we're going. But perhaps also this interim step is being able to do a thorough and very quick analysis, and then serve up recommendations of how best to meet prospects, from a very, very specific sort of like a niche program, as opposed to prospects a little bit more generally. I don't know. It's I don't know if you guys agree with that. But I feel like that, that is a super interesting insight. Like, right now, if folks are going to do any of this analysis, it's, it's a pretty manual process, right? Like, it's really hard to kind of dig into you and, and ascertain trends from how prospects interact, again, at a program-specific level, versus at a school level, let alone just as a, as a graduate school in general level.

Well, I mean, our whole app is basically built to do some of that automation for you, right, so that's what the AI that we've been building over the past five years does, is it actually works around the segmentation and, detecting behavior and, and bubbling that up so you can segment that the part that we don't have yet is the well what do you do? Like, what's the generation part and that's where schools have been like we see a whole huge gap right now is that everybody's really busy. On the other side, we have the technology we are, and we're bubbling up all these segments, but the schools don't have the capacity, the technology, the manpower, the know-how the marketing to TO GET and PUT communication sessions and do reach outs that are personal to the students. Yeah. And that's where we are. So that's the exciting part. The exciting part is that now we can kind of close the loop. Yeah, yeah, we can identify those people. And then we can take action to generate content. That is much easier. So that's, that's pretty exciting.

Yeah, no, very, would you add anything to that Petar?

I would like to add. So what I see that's, that's really, really interesting here. And, as I just mentioned, that's something that we saw over the past few years as people are either too busy as well. But sometimes people are just not marketing savvy. Yeah, or they're not writers, me personally, as well, right? If I were to write an email, or a marketing email and need to attract attention and focus and be something funny and interesting array, I can be hard to see, you're just you just want to get started. And then maybe you can play around after that. Yeah, so these tools will try to save you 90% of the time, it will try to get you there, you will still you can still interact and put your persona into them. But how I envisioned this, it'll be a big time saver, like all over the place, you still need a platform, you will still need a platform to have something to use learn to use, but it should get you 90% of the way there. And then the rest 10% is for you to react. As far as ours is mentioned. We get we got the data, we can surface that. And now is we just need a way to surface reactivity. Yeah, the ability to react up to onto that. And that will be easy, but it still needs to be a personal choice, right? You want to do something or tell them what you need to tell them. Hey, you know, do you think this is what we think for what we see? And these are the options that you can you should do? And this is what we expect to see as a result of those options. Right? What are the possible outcomes? Yeah, they can drive their business, right? Yeah, that's promising, very promising.

Yeah, it's, it's super interesting, because like, like, what you were, I think, both just getting a little while ago is like, who cares if you can do all this segmenting. And you can do all this versioning. And you can, you know, leverage all these incredible insights that like a tool like Element offers, who cares if you still have to do the work of like, then creating 17 different versions of that email? Right? It's like, you have the insight, you know, like, hey, school, you know, if just write in this particular way to this individual Go, Go Come on like we this is how you want to do personalized communications. This is personalized communication. And they agree 1,000% with you, but they're like, but Ardis, I'm, you know, one person or pet or like, I have one and a half staff, like how can I possibly write 17 versions of an email for an open house like no way, Jose, but these tools, if they can do 90% of that lift, and if I just have to go in and like review, and just very quickly, make sure like, you know, Hey, this looks great, or let me add, you know, change a little bit of a missed context here. And that takes me an hour for 17, you know, emails, you sure as heck better believe I'm going to spend that hour doing that. Because the potential outcome of these being highly personalized to these, these segments is so much greater than any outcome with just one version of an email, blast it to all of these segments might be.

And you can also think about the following scenario. There's also a lot of turnover happening. Yeah, there's a lot of auto knowledge that's being lost. But all of that knowledge is probably in some documents, it's available, you know, in certain knowledge bases on FAQs, and websites, but people still need to ingest them. And these writers who will go there to write those emails and put some university-specific information in them, there's no need to learn that the right people need to every university might be different. With these tools, they are able to ingest a lot of content. And depending on a topic, they can semantically find that meaning and generate content on a certain topic. And then you just go in there and adjust the detail or atone. So new people coming to the job can be productive they want, they can send out emails they want, even not knowing a lot about university because they can use all that knowledge that's already built in. So that means that like losing people, right, if they leave, you won't lose the knowledge is still encoded in there, because the AI will keep it retain it and will still keep it give you the suggestions.

So that is a that's yeah, that is a really great insight. Because like, that's something that I wasn't even thinking about, right? It's like yeah, like what is all that historical context that is lost. Or another example too, that just came to mind is imagine right, being able to go into your CRM, like you go into Element, right? And there's a little chat window or whatever and you just say like Hey, what was the average open rate and click-through rate of all the emails that we sent for last fall's open house and you just click go right, and then immediately you get a response of the average open rate was 32%. And the average click-through rate was 7%, or whatever it is, right? And then if you if you're new to the job, you'd be like, Alright, so this is the benchmark, this is the baseline, I'm going to try to like beat this time around and like, you'll be able to know very, very quickly, rather than having to go scroll through and you realize no one followed the naming convention, it's impossible to know what this email was tied to someone hit clone, clone, clone, clone 17 times, and it's the same name of the email. And like, it's just impossible to discern sort of like this, you know, campaign-specific data, unless people you know, religiously followed the setup processes, which most people don't. So that is a fantastic insight pattern and like, something that is super, super interesting to consider, again, in sort of this realm of like savings of time. I'd love to hear if you guys are willing to share a little bit about specifically to Element right, you guys are clearly thinking through these things. And on the forefront of these conversations, which is, again, one of the reasons why we're having this, this four-episode series is because Ardis was like, Hey, we're doing a lot of great stuff here. Like we want to, we want to share this with the greater higher ed community, right? And I was like, of course, like, let's do this, right, and you guys have always you guys are always seen as sort of a, you know, progressive, innovative kind of player in this space. And so it's very appropriate that you guys are leading this conversation as well. So I am curious, like, what, within the context of Element's specifically, are you guys thinking through? What are some things that you're working on? What can you share with us about this, this secret, like, black AI box that you guys are working, that folks might be able to get excited about?

Yeah, I can give you the division. And then Petar can kind of jump into some of the details or specific use cases here. So when you think we've talked about this in the prior conversation, so it's around personalization, right? So we're working on AI around personalization, efficiency, and also access. So those three things are super important, right? So personalization, we talked about, Hey, how can we derive subject lines? How can we derive content that is very personal to that student, some of our email and SMS and kind of campaign tools and Element's and landing pages, like we already have that foundation, and now we're working on kind of generating content on the fly that is personal to that individual that's going that email is going out to and also making recommendations around better subject lines better so Petar can kind of talk a little bit more about that. But it's around this recommendation making things easier. We had introduced, like, you know, a couple of years ago, we introduced this whole notion of a campaign in a box or automated campaigns. Yeah, and that was the vision, right? It's like, Hey, you have a template of this thing. And then you just inject a couple of things that specific to this school, and it will build something for you. Now, imagine that and taking that and now even personalizing everything around that particular communication, or that pack that is that looks and feels very personal to the school. Cool. Yeah, it doesn't look like a template anymore. So our foundational work there is now going to get better and better by adding the, you know, the AI to it. The other thing that we're thinking about is that we think about the human interaction, or the end-user interaction with Element because ultimately, we feel like that Element is a tool that is helping schools to connect better with students, right? And it's all about that connection, and, and students being able to succeed in kind of getting a degree or, you know, kind of getting enrolled to the school or even graduating and kind of communicating better with so in that sense, we build like we focus Elements a lot on the communication part or the experience of that user with the school and the tool. So when you think about that, like we've built you know, great tools there but as we move forward, you need a unique Camino you need something that's a lot easier to interact with and chatbots are really that that tool, but now they have the ability to provide the context in those in those communications so that's something that we're focusing our conversations tool is a multi-channel you know, two-week communication, kind of platform, chat, live chat, two way SMS, all that inbox, but now add on the ability to automatically have the knowledge base and have the communication and interaction with that student, be driven by AI answers, right? So the student has this one pane of glass, which is that Chatbot. And now just like you said, they're asking questions, they're doing work in there, they can sign up for events, they can sign up, they can, they can even complete the applications they can ask for, for grades, they can do all of that thing through this one pane of glass. And behind it is all of the complexity that goes with a tool, like a CRM, because we have the context of everything. So that's the vision, right? The vision is that simplicity should be the key to everything that we're doing. And we're simplifying the content generation. But then we're also simplifying the user interaction and that experience with a student. Yeah. Wow. I mean, Petar can dive a little bit deeper on some of that? Yeah. I'm sure you have other questions as well.

Because I think what Ardis just painted is like sort of this, this beautiful image of like, what everyone in higher ed dreams of right? It's like the simplification of a system that is quite bureaucratic, and honestly, quite complicated and confusing, right? Like, as an industry, the processes that most institutions have, in order to inquire, apply and ultimately enroll, it's just there historically, they historically have been full of friction, right? And so like, what you're painting Ardis is this beautiful vision of like, hey, what happens in the future is that a lot of that friction just gets like reduced, right? Because of these tools. So Petar like, talk to us a little bit more about what Ardis was this vision that Ardis painted? And from a technical perspective, like, how is this going to happen?

Right, so the main important thing is, first, you need to have a platform to solve that problem, even without AI. Right AI is, is a tool that you use, it's not your solution. Right? So we got the platform. Right. So now we're thinking, Okay, what are we offering, right, we have our messenger, we have our chat that our customers use, let's Ardis mentioned, that's a really powerful, powerful delivery mechanism, right? It's something that we initially when we were designing it, we were thinking like, Oh, this is really powerful. This can be a powerhouse of applications. It's not just a single chat, we can power, different widgets, pop-ups, that can show up on the page to enrich the experience for the students. So that was a vision even before AI now without, with all these things, how they progress. It just escalated, and we got really inspired. So how we're thinking is AI will be used to remove the decision trees, right? That's the problem, right? Because you cannot encode them, you need a lot of people and a lot of knowledge. And that's configuration. Nobody wants to work with configuration, right? It quickly goes out of hand. So AI will be used to remove the configuration part. The whole idea is it's it should be as simple as turning on it understands your processes and is able to work. Right. So if you're talking about, for example, a QA chatbot, for example.

Your that is something that we're working on right now. Yeah.

Yeah. So what's the perfect user experience onboarding a chatbot? Right now, well, what are we gonna write down doing, they need to encode all the different knowledge, right, all of that right, then probably set up decision trees, you know, hey, if we detect this intent that goes to this, etc. But what's better? Imagine if you just point to your FAQ document or FAQ page on a website somewhere. And just to keep the Chatbot knows all about that, as if a person read all of that and knows all that and is able to formulate responses, summarize them answer in different tones, and even adapt to the student. That's what's possible now. And that's what we're striving to implement actually. So that's the first. Second is like that, that's again, that that is a helper to boost productivity, right? You don't need to answer now commonly asked questions. You don't need to answer the things that happened during off hours, because probably 90% of them will probably be resolved by the bot itself. But so that we think how can we get further like what's further, can you not interact at all right to skin you can you maybe just leave the bot do everything and do it as if you get that person employed, but at a massive scale. So you will need to work will need to understand the intent and be able to react on that intent with actions. But again, those actions need certainly to be built in the system systems need to support it. But But this is the current industry trend right? The bots or the machine learning models are able to do that they can translate certain intent to an action and then you can hook that up onto your system which can then go further. So that what does that mean? That means that a student can go in ask for their application status right and they can get that immediately, right? They can register for an event, write what is able to capture inputs, understand them, and give it to the system, and system can call that and send it somewhere right. So That's the that's those are the little add details of.

How we think when we talk about. Yeah, so Petar makes a good point. But sometimes it's like, you can pontificate all day long and say, Oh, this is going to be, but when we talk about, like, the things that he's mentioning, the QA stuff, like we're talking, like weeks, and then like short term, medium term, long term, when we talk about that, it's like weeks, a couple of months, and then six months is our long term. Yeah. So you know, over the next six months, like all of the things that we're talking about, they're either, you know, being tested right now in production are in kind of a test phase, we figure out the technology and, and all those pieces. So over the next six months, there's gonna be a huge change of all of these tools is now attached to reduce that work there.

Yeah, no, it's that's it. That's so incredible. And, you know, one of the things patented that you said, which I think is like a really important thing for everyone to remember here, right? It's like this idea of being able to interact as a prospective student with a chatbot. And it's like one interface, and you can ask it, whatever you want to get what you need, right, which is essentially what ChatGPT is right? Like, that is why it's exactly the world by storm is like, it's one interface, you get whatever you need from that. And the reality of the situation is like, even for the people listening right now that are like, Oh, this, you know, yeah, like, We'll see, we'll see about these, these tools and these things, but there is an entire generation right now of your future students who are being introduced to information in the context of ChatGPT. So, like, You, you, this is not like a, okay, maybe this is like a luxury good this, this is a necessity, because your future prospects, the people that you're going to want to attract to your programs, they're not going to search in the traditional way that that I searched for school, right? You guys might have searched for school, you know, and think about people who are 10, 15, 20 years older than me, they all did. They all like to understand a hierarchy and like navigation. And like when they go to a website they go to the nav, and they go to the sub nav, and they find the information that they're looking for because that's how they were used to content being organized. I like never go and sort through like an app bar, I go to the search bar, I always go and try to find the search bar. So I can quickly look for what you know, and find what I'm looking for on that website, right? This next generation of prospective students, they're not even going to search, they're not going to go to the search bar, they're going to expect this interface that they can very quickly and easily interact with in the tone and the style that they prefer. And they're going to expect a really quality answer. And if they don't get that, they're not going to have the patience to stick around. They're gonna go find something else. And like, unfortunately, or fortunately, depending on how you look at it. That is just the reality.

Yep, yeah. One last thing I wanted to add there, it's really it's been really difficult in the past is multi-language translations in terms of QA. And so access to information, we get asked every single day by our partners, hey, can we have multiple, like we've done translations, and we've done internationalization, our applications and all that, however, writing the content is a problem, right? So to have the content in English, and now how do you provide that same answer in, you know, Spanish, or French or whatever, whatever that language is Chinese. So as international students are coming in, they're asking those questions now would have been really nice to answer to take that same knowledge base and, and be able to give that answer in the native language that that person is interacting with. So that is a huge opportunity that is going to be you know, that is going to be delivered. Because we're getting that for free, basically, because of these large language models. And we're basing our AI on our conversational AI on the same model that ChatGPT is built on. Yes, it is probably getting too technical, but it's in the open AI. It's an open AI model that is the same one that ChatGPT uses. So you get translation for free. So now you can say, give me this answer in Spanish, and it will actually translate that same answer to you in Spanish. So you don't need to have multilingual knowledge bases anymore, or translate your website because that's all going to be done by the model.

Yeah, that's the second component, right? As you mentioned, like there still needs to be a language knowledge base right there. doesn't need to be somewhere it can be written in a single language right? The bot needs to be able to search for it. Right. And that's also one of the features that machine learning models give us is that semantic search. Yeah, people are mostly familiar with keyword searches, you search for something, it needs to match a certain keyword. That's impossible, right to match different languages like you've like us. But if you think about semantics, what it does is translate the meaning and is able to encode that meaning and search by meaning. I mean, we can go into details about how that actually works.

I mean, it's very generic right now. Right? When you think about ChatGPT, btw, you can go in there and say, What, When? When is the school closed? Right? It doesn't know that. Because it doesn't have the context. It just has general information. So it's going to give you an answer, that is not even correct. So we got to tell it, it's like, Hey, if you don't know it, don't give us an answer. But now what we're able to do is we're able to give it that context. But in order to pull that context from hundreds of 1000s of pieces of text, that might be our website, FAFSA, you know, like government data, you know, all of these things that are very specific to higher education, they might be proprietary, or now you need to be able to search to give the context because you can only give it so much context, right, you can only allow so much context. So you need to be able to give it the right piece of context. So semantic search around that context becomes super important, right? And that's kind of what Petra is talking about, is they need to be able to find that content that is relevant to the question that's being asked, and then inject it into the prom. So then you can get the right information. And the bot can kind of give you the information. So there's a lot of steps that go into that part.

And think about all the variations like the same the question can be in different languages you have, right, exactly. The content itself can be in different languages. And also the answer should be in a different language and different tones. And all that is not possible to be accomplished quite easily using these machine learning models. So that's, that's the wow.

Yeah, yeah, it is a well, and on this note, too, right, like, as an industry or like, like higher ed considers itself to be a very accessible, right? Exactly. Industry, right? And accessibility is incredibly important to higher education as it should be. And also, like higher ed has a very, most higher education institutions have a relatively diverse, quote, unquote, customer base, right, like, not every organization or company has, as dramatic of a customer base as higher education institutions do, right? Most higher education institutions have several people, right, like several meaningful percentages of their students coming from other countries, and not just one other country, but countries all around the world. And so if for no other reason, to get excited, and pay attention to what's happening right now, in AI, school should care purely for the accessibility components that AI can offer. Even if they're having issues with everything else. Fine, ignore all these other things. But where you guys are both just with where you guys are both hitting it right here is so so important that it actually makes accessibility possible, in a way that historically hasn't been able to be possible, unless you could afford to have staff who were well equipped to translate every single communication into every possible language that your prospective students might affiliate with.

Exactly, exactly. And the student can answer can ask questions in their own native language, rather than figure out how to ask it in English and then translated or looking for information. So it's, it's, yeah.

Wow. I mean, I'm so excited guys, for what you all are doing an Element here. And I think that you all kind of being the pioneers in this space of figuring out alright, how do we take all these things that we hear about ChaCha PT, opening? I, you know, the future of machine learning these, again, buzzworthy terms that everyone's talking about? How do we like work these into a meaningful context that serves us as higher education, marketers, and admissions professionals in more concrete ways? And so I'm just very thankful for all the work all the time and energy and r&d and money that you guys are spending to try to figure all this stuff out. It's incredibly important for not just Element partners, but the greater higher education community to be able to look to as an example of people kind of doing this work for folks that do want to learn more, or kind of just stay up to date on what you guys are building. What's the best way for them to kind of like get in touch?

Yeah, you can just go to our website element451.com and then, you can kind of look through our resources. We have a lot of resources there and we're posting blog posts all the time. We also have a conference at the end of June called Engage Summit. And that's where we're going to make available a lot of this new technology that we're talking about obviously before that as well, but that would be a really good place to come in and build community, you don't have to be an Element customer. You know, it's here in Raleigh, North Carolina, and it's a place where you can get a lot of ideas. I know, Zach, you're one of the participants, one of the speakers last year at that conference, and everybody loved your session around email marketing, and kind of what drives students and so that was a popular one, but it's a great place to go.

Yeah, yeah, I can't recommend the event enough. I think it's really cool to see you guys taking this position in the space of being obviously a software provider but creating an event that's not just a user conference, there's you know, several others kind of like in the space that creates essentially like user conferences, but it's cool to have you guys pioneering again an event that's not just an Element user conference, but really sort of a broader sort of an educational event for people who are interested in understanding sort of the future of marketing and technology and admissions in higher ed so can't recommend the event enough and again, we'll have a link in the show notes below. So you can scroll down and just head on over to the event landing page if you want more information there but Petar and Ardis thank you both so much for your time thank you for the again the innovation you guys are doing in this space super grateful and thankful to Element for helping make this whole mini podcast series possible and again, if you're just tuning into this episode, and you haven't listened to the previous three episodes, click on episodes one, two and three and enjoy. But thank you, gentlemen, for your time.

Thank you. Thank you, Zach. It was a blast.

When it comes to the student experience, we know that you want to be a trusted guide from recruiting to graduation. The Engage Summit brings the best minds in higher ed together to give you the strategy and tools to create a cohesive student experience — from start to finish.

Explore the latest technologies, increase your skillset, and gain insights into today’s students, to deliver the most personalized digital engagement experience every step of the way.

This is not your standard edtech user conference — this is a dynamic, inspiring, and empowering event for all higher ed marketers and admissions professionals.

Explore the latest technologies, increase your skillset, and gain insights into today’s students, to deliver the most personalized digital engagement experience every step of the way. June 27-28 in Raleigh, NC.

How to Use AI to Analyze Your Data and Get More out of Your Higher Ed CRM

Related Articles

Best practices
How Higher Ed Can Leverage ChatGPT for Marketing and Student Recruitment

Best practices
What Higher Ed Marketers Can Learn From How the World's Best Agencies are Leveraging AI

Talk With Us

At Element451 we help schools with software that keeps students engaged and enrolled.