Novakid Launches NovaPals for AI-Powered, Conversation-Based Language Learning | GSVtv Podcast

Speaker A: Foreign.

Speaker B: We are here at ASU GSV Summit 2026 with a really exciting product announcement from NovaKid. We're here with Max Azerov, CEO and co founder of Novakit and Chris Kelmsley, who is a product lead for Nova Pals, which we're just releasing today. Um, let me first tell you a little bit about them and then we'll jump into the questions. Max Azeroff, experienced entrepreneur and one of the two brilliant minds behind Novakid, an award winning online English school for young learners. Max has a passion for technology and innovation reflected in his impressive list of startups and NovaKid's success as well as his ongoing commitment to professional development and business transformation. Chris Kemsley is a product leader building AI first apps at NovaKid. He specializes in OTA one, uh, zero to one digital products with a track record in educational platforms and generative AI for language learning. Today, Chris leads Nova Pals, an AI English language learning app launching at ASU GSV 2026. Nova Pals helps kids go beyond classroom English with AI companions they can practice speaking with anytime. No tests, no pressure, no parent supervision required.

Speaker C: Welcome to the podcast.

Speaker D: Thank you, Alex. Nice to be here.

Speaker C: Yeah. So before we get into Nova Pals, I know that's the big news of the day, let's talk about Novakid itself. Tell us about the origins of Novakid and how it's grown quite big.

Speaker D: Yeah. So, uh, definitely goes from the personal, uh, frustration to a large degree. Uh, I grew up in Eastern, uh, Europe and I was learning English in public school and definitely had this firsthand experience where, you know, the teachers that we had were not really all that fluent in English to begin with. Uh, so we did not get a lot of practice, to be honest with you. We get zero practice. So we were studying English as a sort of like theoretical, abstract entity, not, not like a living language that it is. And this was the kind of, the insight and the frustration, uh, that led to novakit as a way to give kids simple, um, access to, uh, fluent tutors, fluent teachers who could help them practice the language as well as practice with peers from other countries.

Speaker C: Yeah. And you are now in over 50 countries, hundreds of thousands of students and

Speaker B: you're using AI led learning and practice

Speaker C: with reinforcement and one on one live teacher classes. Tell us what the experience is like for a student Novakid today and then we'll look into what the future of Novakid looks like.

Speaker D: Sure. So, uh, historically we started from uh, one on one classes with live teachers, uh, and uh, this was a hit product for us for a long time. But we're stepping into the new era believe it or not. And uh, yeah, definitely the experience changes and uh, we adding uh, a lot of AI components into the product. Uh, AI tutor has been an important part of that but this is not the only part. So uh, we also have LLM generated content that students, students are uh, experiencing and uh, improving their uh, test, uh, their training, training experience within the product. Uh, we introducing a hybrid approach where you have AI and you have group practice experiences with live teachers. So basically you kind of practice, you train with AI and then you bring this to a group in order to actually practice with living human beings. Which we find uh, students enjoy but also parents appreciate a lot. So those are kind of the main ways that you experience AI group experiences and one on one learning, um, within Novakit.

Speaker C: Yeah, that's very exciting. And you have tons of practice. It's beyond so different than traditional language learning where it's very theoretical. It's all about practice.

Speaker D: Yeah, absolutely. We're really very practice heavy.

Speaker C: Yeah. And now we are launching Nova Pals. So Chris, tell us about Nova Pals. You're the product lead for Gamified Education. Give us the overview.

Speaker A: So uh, what we're doing is we're taking kind of seven, eight years of Novakit's deep learning expertise and curriculum design. Uh, we're repackaging it as a practice app for Tease. Yeah, um, it's got gamified features in it from zero. We're not kind of bolting on streaks to existing product, just trying to make it exciting. Uh, and it's based around a kind of core loop of talking to these engaging, team focused characters that like stuff that kids tell us they like music, art, uh, gaming, uh, and playing fun games toward quads. So it's super fast, super fun.

Speaker D: Uh, and we're just trying to aim

Speaker A: it at a market where kids want to practice in there. They don't necessarily pay for a main one to one tutoring product, but they're learning English in school and they're of deserved at the moment to have a load of practice that repetitive.

Speaker C: And you mentioned the core loop of the game is about conversing with AI characters. Tell us what that actually looks like in practice. Ah, what is the student doing on their phone in a daily basis.

Speaker A: So we're taking our kind of design cues more from gaming ads. We know where teens spend their time.

Speaker D: Right.

Speaker A: It's Fortnite, uh, it's Brawl Stars, uh, that kind of look, uh, we're designing from that principle first, right? Yes. So it doesn't look like the normal ed tech pack. It doesn't look like normal EdTech school textbook characters.

Speaker D: Right.

Speaker A: Uh, we know this market are suspicious of stuff that adults have designed where they don't think it's for them. So we've taken great care to try and do that as a first principle.

Speaker D: Right.

Speaker A: Um, you're having these small chats. You have goals you'd have to achieve throughout the chat. Um, there's kind of a discrete assessment tool so that we know you hit those goals and you use that to win points, customize your character, unlock blue devils, all the stuff you're not expecting.

Speaker C: Max, you've mentioned that you're really focused on underserved learners. There's certain ages of learners from language that are really underserved and there's a sort of age gap where they get it at a young age and then revisit language learning older. But as a teenager, especially when they really, really need to understand a new language, they are not always motivated. How are you thinking about that and how have you taken the data from your existing Novokid? Many, uh, years of teaching a Novokid and instilled it into this new product?

Speaker D: Yeah. So definitely, I would say for us audience it's a, uh, the easy way to explain it is, uh, it's very similar to math learning gap. In the US I would say worldwide there is an English learning gap that exists and for already mentioned reasons, the lack of uh, fluent tutors, uh, the lack of practice, uh, opportunity and so on. So this was the initial insight. And uh, definitely when you get exposed to a foreign language, uh, from like between ages 6 and 7, it definitely helps a lot to build up the confidence early on. And especially for teens it becomes a difficult, you know, as you get older it becomes more difficult to kind of break through this barrier. Uh, and we find that, you know, the earlier you start, the better.

Speaker C: Yes.

Speaker D: So obviously for teens it gets, you know, we get a bit of a higher barrier to overcome. But you know, having this really cool character, very quirky characters, it really does help to kind of, you know, break the ice and get the conversation flowing.

Speaker C: Yeah, you mentioned sort of Brawl Stars and some really entertainment based apps as inspirations for this Fortnite. Um, you haven't mentioned Duolingo, but I know that duolingo is on your mind as you think about gamified learning. And how are you going to take on the behemoth of duolingo for gamified learning and displace Them, Uh, well, I

Speaker D: don't know if you displaced them, but we definitely want to uh, make a better product that is specific, specifically designed for young learners.

Speaker C: Yes.

Speaker D: And we believe that you know, by essentially using this AI first design compared to Duolingo, where you still have this more of like a click type of uh, uh, practice as a first, uh, and then AI sort of comes the second. So we think that this audience will actually be more receptive towards this. And uh, this is the bet. I mean hopefully we get it right. Hopefully it will uh, you know, if the teams like it, if the young learners like it, then it's just going to happen naturally.

Speaker C: That's right, yeah. And when you think about that core loop and moving beyond the sort of click to more of a conversational, super hyper motivational, uh, design as the product lead, um, how do you test it with kids? How are you thinking about getting feedback from your first audience and iterating into something that's going to be really change the market?

Speaker A: Yeah, I mean we've obviously got a lot of analytics in yam, so we tend to learn by just launching it. Like you know, we are ah, officially launching today, but we've got 15k plus users, we're in a couple of test markets. So we've been having early data for about a month and a half, two months now. Yeah, um, so we're mostly learning that way. But they're putting it in kids hands. Everyone here that the kids just put it in their kids hands.

Speaker C: Right.

Speaker A: Almost seeing what they want to do while they don't.

Speaker D: Yeah, they just use those as guinea pigs. That's our kids as guinea pigs. That's what happens.

Speaker C: And you've gotten feedback from many learners in your existing network. I mean you're in 50 countries, you have so many hundred thousands of kids. I'm curious what lessons you've learned from the core novakit product that are being instilled into Nova Pals.

Speaker D: Yeah, we actually started experimenting with AI a little bit, uh, like last year I think. And uh, one of the first experiments was this cute little panda, uh, that just suddenly appeared in uh, Novakit, uh, experience. And what I learned, uh, what surprised us is that kids actually uh, really love to speak to the characters. They really make emotional connection, uh, they make plans for the next lesson like hey Pandy, what are we going to talk about? Are we going to go football like next time? So it was really surprising that it was going relatively deep. Uh, but you know, on the other hand, maybe not so surprising, you know, you know, my kid would probably prefer speaking to a pandy than to a middle aged guy. So I guess in hindsight maybe that's so surprising. Uh, and uh, yeah, another thing was surprising is that the speaking time compared to the uh, overall session time was actually much higher in uh, AI versus the live teacher.

Speaker C: That's interesting.

Speaker D: So almost twice, twice as high. But you know, on the flip side, the actual sessions are quite a bit shorter with AI. So if you, you know, this is the nuance. Right. The speaking time is pretty high. But if you want high task at hand, then live teachers are actually better at kind of keeping teach, keeping kids motivated and kind of staying sticking to the nitty gritty of things. So that led us to believe that you know, ultimately there's probably going to be like a spectrum of learners. So more of like casual learners will probably be okay with AI only uh, learning experience. But somebody who has a bit of a more uh, uh, ambition to it would still want to have some uh, human component.

Speaker A: This is why we're both designing both

Speaker D: products, both AI first and uh, hybrid products as well. Yeah.

Speaker C: I think you're homing in on an incredibly important aspect of this sort of AI in education moment, which is that we believe that AI can do education really well in certain circumstances, in certain ways. Nobody thinks it's going to replace teachers outright. That's just not possible. But it can do certain things really well. And the idea that you're getting more speaking time but less focused speaking time on the exact task at hand, that makes sense. I think it resonates with what we think about when students use AI. They can sort of take it in any direction and it'll follow them. But that's a little different than what a human tutor will do. Some combination is going to be really interesting. How do you think about the combination of human tutoring or human teaching and AI teaching to get sort of the best possible outcomes?

Speaker A: Well, we still think of Huey just predominantly doing the teaching, AI doing the practice.

Speaker C: I see.

Speaker A: Um, we've tried both doing both. It's not a great use of teachers time. Having them doing repetitive drilling over and over and over again on a video versus picking up a fun app that makes bleeps and blues. You're having fun and doing the app work there, right?

Speaker C: Yeah.

Speaker A: Um, I guess the interesting thing building the other way around is we're trying to create these characters that are super engaging and fun and you know, we're calling them Nova Pals.

Speaker C: Right.

Speaker A: Uh, but we don't want to take that too far. Right. Like we Wanted to know your name and your hobbies and like what you did last time and what you'll be doing next time. But we don't want to remember like your full name and your address and your bank details.

Speaker D: Of course.

Speaker A: It's this like we've got guardrails in there to keep you super safe. Right. It can't place a teacher. It doesn't have that emotionally intelligence and we don't want it to if it's not right. Trying to do.

Speaker C: Right. So this Pandy character that you're in, tell us about how you came up with it. And as you've been testing it, you're

Speaker B: seeing some really intriguing numbers.

Speaker A: Right.

Speaker B: With over 60% of beginners on the platform actively using it, over 15,000 active

Speaker C: learners, as you said. What did building and teaching this Pandy character really teach you about the design of the app and about as you say, keeping those guardrails, but also keep building a meaningful relationship so that you could say, I'm so excited to talk to you next week, Randy. Like there's an interesting balance to strike there. I'm curious how you're thinking about it.

Speaker D: Yeah, I think the main thing it taught uh, us is that uh, it actually works and it works to a degree. Uh, and this is actually what gave us the confidence to go all in on this AI first product which Chris is leading.

Speaker A: Uh, sure, yeah, I think with um, with that one again, we shipped it super early. We started gathering data, we started it on younger lovers. So like our new products for teens, uh, it made sense for the first one to keep it super simpler. Their language is super simpler. The quirky has an easy task to design an AI for. Um, but yeah, we saw pretty early that a, uh, large percentage of our users took advantage of it as a practice tool. Uh, I think, yeah, whatever that number is, 60 something percent of those, 50% of them use it twice a week at least really for between two and four minutes.

Speaker B: Wow.

Speaker A: Around 10% of your users use it five times a week. So you're probably kind of daily now here, but they'll go and do that lesson with the tutor. Five days with the panda, back to their real tutor.

Speaker C: Yeah, that's really interesting. Yeah. And those usage numbers are really powerful. How are you thinking about measuring, you know, measuring deeper learning over time? That's always the million dollar question with anything in ed tech. How are you thinking about where it's going to go and uh, are you going to use external assessments or anything else to make sure that students are really absorbing that language? This is something, uh, other language apps have been accused of not paying enough attention to.

Speaker D: Oh, we definitely are, uh, taking advantage of some of the, uh, you know, standing on the shoulders of the giants, so to speak.

Speaker B: Yes.

Speaker D: Right. So, uh, there is this Sephir framework that is commonly used in Europe that basically, uh, you know, the level of proficiency in language, not just English, any language and spoken written language. And we're taking advantage of that to basically score how kids perform, uh, and also give feedback to the parent that where does your kid stand in terms of the Sephora proficiency? AI is actually extremely helpful on the assessment side of things because it really does help to uh, score, uh, not just the, um, we can literally score just the videos of the uh, tutoring sessions and have a really good idea, really precisely zoom in on the uh, on their proficiency level.

Speaker C: Yeah. And for those who may not know the Sephir platform or the framework, I know it's sort of central to the curriculum you've designed. Tell us a little bit more about it just so that anybody who's not familiar with it can sort of understand this underlying framework and how you're using it to really boost the curriculum right off the bat.

Speaker D: It's basically, uh, requirements, a set of requirements that shows you, you know, what level of the language you are. Right. And you start from a one, uh, then you go to a two and progress all the way to C. C2 I think, is the like, fully proficient, um, so just basically a way to measure yourself against the language, uh, proficiency that is commonly used, uh, across Europe and much uh, wider than that.

Speaker C: Right. So by using that framework you can actually see students making their progress and amendments.

Speaker D: Absolutely. And we don't have to be kind of tied into uh, proprietary systems that are, you know, offered by some of the companies in the market. So this is, seems to be a very good, uh, kind of, uh, more open source, if you will, way, uh, to approach it.

Speaker C: That makes sense. So this is a question for both of you. You know you've mentioned that Novocode took a calculated risk to lead with AI. Right. You said you're being AI native, you're really leaning into it, but that AI

Speaker B: is not about replacing teachers, but about unlocking what matters most in language learning, which is practice.

Speaker C: That's where you started the whole conversation in class.

Speaker B: People just don't get enough practice. With everything you've built and seen, what can a language learner do today?

Speaker C: That wasn't even possible three years ago because I feel like AI is really transforming the language space. I know you're on a panel about exactly that at the summit here. Um, a question for both of you, but let's start with you, Chris.

Speaker A: Uh, sure, yeah. I mean, yeah. What's that? Three years ago so we just all learned about LLM chat.

Speaker C: Exactly.

Speaker A: So none of this was possible as the answer somebody really briefly. Yeah. Um, I guess we've seen certainly in the first stage of that it being applied to every application. Right. And some of them very successfully, some of them much less successfully and you don't hear so much about them. Now we think language practice is a uniquely good way of using this as a tool. Um, do we think it can do everything? No. So we still think we need teachers for all but there's so much scope where we can go with the bits it is good at.

Speaker D: Yeah.

Speaker C: So uh, what types of functionality are you most excited about that it can do? I know it can follow a student into their interests, you can build relationships, you can assess. But I'd love to hear you just because I agree with both of you. I think language is a uniquely, it's a really, really excellent use case for AI because of the practice, because of the frameworks. Um, and I'm curious how you've seen that play out over the last three years because you've. Novokid's been a leader in the space for quite a while.

Speaker D: Yeah. Actually when we just were starting the company my initial thought was to leverage technology and really deliver this ah, practice in a scalable way. And unfortunately after some kind of prodding around what I realized that there is no technology that is good in it. That wasn't when we started that. So the uh, speech recognition was not there to understand somebody who's not. It could barely understand people who properly spoke the language but people who just learning the language. There was just no way this was possible. Uh, there wasn't any intelligence. So even if you do understand what was said, there wasn't really a good way to build the intelligent bot into that. So now obviously all of that has changed and I think from the consumer point of view the main thing that has been enabled is basically on demand access. So one thing that is great about AI is on demand access. One thing that is kind of sucks about life, teachers, uh, with all the good things about them is that they are not as accessible. So you need to pre schedule ahead of time. Um, sometimes they get sick, you know, sometimes they have different moods, uh, etc. But so from the accessibility point of view this is a real big changer. That means that the frequency of Practice suddenly becomes a truly daily habit versus something that just you do like twice a week and even sometimes not that because as I said, teachers are life humans. They do get sick sometimes and so on.

Speaker C: Yeah. Technology is always on, always accessible and it doesn't get into a bad mood and refuse to teach you any new.

Speaker D: Sometimes it does. Some of our characters get in a pretty bad mood sometimes.

Speaker C: Really?

Speaker A: No.

Speaker C: I want to hear more about that.

Speaker A: How do they uh, I mean we haven't tested it. Usually our ah, hunches that strong characters are more engaging than boring characters.

Speaker C: Interesting.

Speaker A: Uh, and what we've seen a lot of in uh, the last couple years are ah, white boring niches.

Speaker C: Yeah.

Speaker A: So we're leveraging a little bit more.

Speaker C: So Vin personalities. Yeah, there he is.

Speaker A: We found it fun internally in test flu, so we started shipping it and we've already seen kind of our um, more engaging characters having better rotation that that.

Speaker C: You know, I wouldn't have thought about it that way, but that m resonates. And the strong characters probably are also moodier characters maybe.

Speaker D: Yeah, they're moodier. Obviously, you know, we have to kind of walk the line to make sure they don't overly moody. But you know, definitely, especially for kids who are older, like teenagers, they, they definitely. They want to see some real emotion. They want to see some real attraction, not just some kind of your vanilla, uh, you know, speech. So they want some.

Speaker C: That makes sense.

Speaker D: They want some emotions.

Speaker B: I like that.

Speaker C: A lot of companies in EdTech see AI actually as a sort of existential threat for some of the reasons that you're mentioning there. It's. It takes some of the traditional versions of ed tech where you know, drill and kill or lots of questions or you uh, know and turns it on its head, makes things very customizable, very accessible, but also you know, just sort of moves attention quite away from the educational use case in some cases. Uh. How do you see AI changing the edtech field for the better? Obviously you're doing it in your products, but how else?

Speaker D: Well, we definitely. So the way I'm trying to make sure my team sees it and the overall company sees it is we look at AI as a technological uh, tailwind for another kid. And to be honest with you, for uh, education at large, uh, for us specifically, there's just so many things that we do better. And this actually reflects the numbers. Our investors are super happy. We have been accelerating our revenue growth for three years in a row and we expect to do it again this year. And in a large part thanks to AI, uh, things like marketing, things like sales, things like customer, uh, support. All of this. Things like marketing, sales, customer support, all of this gets so much more efficient, uh, with AI in it. And this really helps a lot. And I'm hoping that not just NetworkIT, but like all tech companies really find a way to leverage that. Stop looking at AI as a foe and really embrace it and use it as a growth lever.

Speaker C: Yeah, it's great. Uh, anything to add to those, Chris?

Speaker A: Uh, uh, from a product perspective, it's the most exciting, fun tool that's out of the Marines. Like, uh, as a thing to build features on. It's a theme that assists with my job every single day. Similarly, you know, trying acquisition experiments using AI generated, uh, adverts and AI jrs complex.

Speaker C: Yeah.

Speaker A: That I can launch myself onto a landing page I build myself.

Speaker C: Yeah.

Speaker A: With, uh, video adverts I build myself.

Speaker D: Yeah.

Speaker A: I can do it in about two hours. Since it was launched, like this job I think has never been, uh, more fun than this.

Speaker D: Yeah. Guys, we're such exciting time. I remember first time I actually fired off, uh, ChatGPT 3.5.

Speaker C: Yeah.

Speaker D: And I, uh, don't remember what I asked. And when it came back to me, I was like, this cannot be real.

Speaker C: Yeah.

Speaker D: So, yeah, this is totally, absolutely out of this world experience. So guys. Yeah. Exciting, uh, times.

Speaker C: Right. And I noticed that some of the things you're recommending are not education specific. It's any company can get marketing and sales and customer success and AB testing or product experimentation. Marketing, testing, acquisition. Um, so that it is, it's really exciting for EdTech. But you also mentioned like content creation through LLMs. Um, absolutely.

Speaker D: Yeah. Content is a big lever for sure for us.

Speaker C: Yeah. And I mean that's not a bad thing that this is. It's sort of lifting the entire business and software field. So just as we close out, you know, Nova Pals is about pals. Right. It's about creating relationships, sort of pseudo relationships. Uh, it makes me think of all the different types of relationships we've had with fictional characters. Relationships with like Tamagotchis in the past. I mean there's so many different ways we've sort of related to fictional characters within toys, within learning. Um, when you sort of think about your most exciting vision of where Nova Pals goes, what is the relationship between a kid and their Nova power? What would be the dream vision for that?

Speaker D: And sure.

Speaker A: And again, we've got to be careful. Like, this is more on the Tamagotchi perspective that uh, it's pretending to be your best friend to. It's a fun thing that you use for entertainment. Not like a bad thing that good cause harm. Right, Right. Like, I don't want to dwell on this too much, but like we have to be conscious, of course, as uh, builders. Um, but doesn't mean it can't be emotional or it feel, uh, in a way more magical than just a robot. Right at the moment we're talking about a narrative of a new character where they start being white sad in any shell. And over five lessons you bring them out of their shell and you've found something that makes them M have.

Speaker D: That's really interesting and we think we can build that.

Speaker A: That's a super exciting thing. Um, so that's kind of where I'm interested in. I think there's storytelling or narratives we could do that kind of hit the same kind of things as watching Pixar 4. Like, uh, a good piece of fiction. That's kind of where I'm.

Speaker C: I love that. Really interesting. So congratulations on launching Nova Pals at the ACGSV summit. We can't wait to see what's going to happen with it. I'm very bullish on relationship based AI uh, especially for language learning. So I think you guys are doing something incredibly important. M thanks to Max Azrobe and Chris Hemsley from novakhid.