You know, the Internet is basically gonna move to this agentic or at least dynamic experience. Everybody's moving to AI for their search and discovery, but why is that the case? It's the case because it's just a better experience. The reason that we're not seeing it everywhere already is in part because the technology still has to basically distribute and and sort of get good at all these different use cases, but also in part because it's really expensive to scale. Does that mean that there there's no reason for me to go to a website? Is it gonna be where in the future I'm asking it whether that's voice or whatever. And it is creating a web experience for me that is curated to whatever I'm looking for. Alright. We are here with Nick Baird, cofounder of Koalabs. Welcome. Thanks, Keith. Okay. First question. What is it like to be starting a company that could objectively just piss people off that it even exists? Yeah. People the general sentiment around ads is very bad, and I think one of the core theses that we have at Koa is we're trying to change that. Yeah. We try to be very, very relevant. It's part of the core offering that we have. The formats that we have are beautiful, and in many ways can be a better experience than the publishers to absolutely work with themselves. And that's a core part of, like, what we measure in terms of success. So as a result, we've gotten a lot of sort of folks DMing us on Twitter or sending us emails, kind of asking for us to bring ads to their favorite coding agent or their favorite AI, you know, travel planner or their favorite AI app because they're tired of spending twenty or thirty dollars a month for something that really can be free, and they see the experience that Azure are are like on other publishers, and they think it's great. And so yeah. So I guess that becomes a question. If, you know, apps, LLMs, what whatever we're talking about here are running ads, meaning they are disclosing that it's an ad in some respect, you would think, and, it is aligned with the intent of the search. Right? Just like Google Ads and all that would be. Right? People know what they expect it. If they are delivering those, do you think this will be a reduction in the amount of money that the consumer pays for that tool, or do you think it will be an offset of the compute cost that they're probably a bit underwater on today? Yeah. I mean, I think it's a little bit of both. There's a lot of different ways that I could kinda take this. The the main one is that it costs an AI app about two or three dollars per thousand queries that that that they process. Right? And that basically becomes prohibitively expensive at scale. So what ends up happening for most folks that are trying to force the subscription only model is they gatekeep all of their best features. Right? And so what we're trying to do is avoid that by having a very relevant sort of ad profile that is also, you know, something that's allowing the publishers to make some money and and and become sustainable, and, you know, in all cases, we're unit economic positive against the the the token costs. And so that's what publishers generally work with us for for a primary reason for, is that they can basically make all their features free and and expand this to an infinite number of users because they know on a unit economic basis they're they're they're net positive. How has it been introduced? And I think to to a publisher, I can imagine introducing this would be, like, very welcomed in most cases. Have you had pub publishers where it hasn't been welcomed? Not with us. I think, as you might imagine, publishers that work with us decide that they wanted to create. Right? And then and then, you know, it's it's something they've asked their users for. And so in in some cases, you know, we work with, like, sustainable model ad. Right? They're like, hey. We offset all of the compute costs and the sort of energy that we're creating, and we go and buy carbon credits for that, and that's sort of their differentiator. And they literally just emailed all their users, and they were like, hey, we're thinking about bringing ads and reducing, you know, getting rid of the five dollar a month subscription, and they were like, ninety nine percent of respondents were like, yeah, you know. That's cool. And so we have never seen that. There have been cases where this has not been great. Right? And then the experience is very bad. I mean, Character Eye is a is a classic one. You can see that they have on Reddit. They sort of had, like, some some issues when they started to show interstitial video ads in between sort of, like, character switches. Right? And people thought that that was very disruptive to the experience, and I think especially on places where privacy becomes a huge question. Folks are, like, concerned about, you know, reading the contents of the message. That is much less the case in productivity tools, which is typically where where we work because, you know, one, there's less privacy concerns, and two, you know, being able to read the contents of the query actually allows us to give a very strong recommendation, and in some cases, solve people's problems. So we've never seen really any, like, negative reaction. So today, somebody goes to a claw to OpenAI, whatever it is, and they they pop into ChatGPT and they search for something. Hey. You know, I just okay. I have a juicer. I saw this juicer on Amazon. This is true. I wanted to get back into juices. I said juicer on Amazon. It's, a hundred fifty bucks, and all the other ones are, like, five, six hundred dollars. Yep. Something's gotta be wrong with it. Yeah. Right? So I asked, but, hey. I do deep research. Do ever find this? Came back, and I was like, nope. It's good. Nothing scammy about it or anything. Was like, cool. So and I bought it. So how do you prevent trust eroding that that isn't a paid, you know, sponsorship that they haven't paid the LLMs to say, or they or they haven't had some kind of bidding feature that the LLMs are like, no, let's promote this. Yeah. I mean, that's a really nuanced question we can talk about forever, but I would say the first thing I would say is I I would challenge you in any way who's thinking about this to say that's not already happening. You know, the land of GEO or AEO exists, and advertisers can pay companies to give them analytics on what organic things they can include in their websites to make sure that they sort of show up. Right? And this has existed across Google forever as well. The best way to think about this though is how you not erode trust is you just deliver something really, really relevant. So example would be nobody thinks that Google search ads makes the Google search experience worse. Why? Because the ads are always extremely relevant to the search. Right? Yeah. So it's kind of a simple answer, but it's true that the way that you don't ever trust is that you make sure that the ads themselves are very high value. What about the visibility of them? I think most people, when they think about ads in these, you know, in native search or or, you know, AI results that they might get back would imagine them being either a display ad or, you know, or it's a pop up, or is it in the actual response and it says sponsored. Yep. Like, what are the different formats of of ads? Yeah. You can do it in a lot of different ways, and the other thing to think about is and we can talk to this I'm sure we get this later in terms of the vision for how we think see the world going, but we actually see this going beyond the sort of chat experience. We actually think the chat UI is sort of a primitive for what agents look like. And eventually, this distributes across the entire Internet. The entire Internet becomes an agentic surface area. Right? And so what happens when that's the case? Well, the formats are gonna look like a million different things. It could look like something that's in line. We haven't done success in that so far. We've really done a sort of native display format, so it is a, you know, a full sort of HTML snippet or a block that we can kind of do our own custom CSS styling in and have buttons and any kind of animations or text or whatever. But even if you think about going on to, you know, say like an education app like Quizlet, right, that doesn't have a conversational interface, while you still might be able to allow a user to go back and forth with a format that maybe is, you know, Harvard's master's degree program trying to get you to be excited about that, and you can ask questions about, like, well, tell me about the professors, and like, how much does it cost, and Correct. You know, like, what is the right are they the best program if I'm thinking about, like, environmental, you know, science. Right? And that format is something that the user can interact with. Right? And so there are a lot of different ways that this can look, and I would actually say we believe that the Internet will become a fully dynamic interface, you know, long term, and so I don't see any reason why ad formats won't also be the same. What about the the display, like, creative side? The I'm the advertiser. Right? Do I have to just give you a packaged HTML five format again and allow you to run that, or can you kinda dynamically generate an ad based on what they're searching or Yep. You know, something that's relevant to them? Yeah. So we already do dynamically create the creative. Right? So we ask for no visual creative for advertisers at all. In some cases, we do take static images and things like that when they wanna have you know, brand control over the asset, and we generate all of the copy as well. So we generate the way that this sort of ad is presented to the user. And we do this, you know, not completely, like, out of the blue. Right? We take the product's features, description, we take the brand and what the brand stands for and who their audience is and all that different information that advertisers know to use in their existing creative, but then we create all the creative on our side. That's cool. What about click through rate? Have you seen that perform? Yeah. So generally, you know, because we're personalizing and because the formats are interactive, we see click through rate pretty high. So on mobile, we've, know, in the early days, we saw as high as fourteen, fifteen percent CTR. Wow. Now generally, our mobile click through rate is floating anywhere between like three and five percent. And on web, it's obviously lower. It's gonna be like a smaller space, but usually, we're seeing somewhere between one and two percent across most of our pubs, which is, you know, very high relative to most of our ad formats. Very high relative. That's awesome. So, okay, what about you've got like you mentioned, there's agencies out there that do GEO, AAO, whatever you wanna call it. Are they are I mean, those agencies I mean, it's like I think of this as similar to the digital marketing world, I don't know, fifteen years ago, whatever it was, and you've got companies that are like, oh, you need a website. Yep. And then companies are like, oh, great. They do the website Yeah. And then, okay, well, you gotta that you gotta have it show up. And remember, like, keyword stuffing and, you know, there's white backspace, the white keywords in there to, like, stuff it. You've got companies now focused on geo. Do you see that evolving to be, like, the next another service of digital marketing agencies? Like, that I mean, obviously, pipeline is not a problem for you, but, like, you get in with the digital marketing agencies, this is gonna be a service they could bolt right on. Totally. A hundred percent. And for what it's worth, we also love a lot of the GEO companies and are pretty good friends with a few of the folks that are sort of leaders in this space, and so we definitely see long term partnership opportunities with them. And, yeah, the agencies of the of that world are really interesting for us. There are lots of different types of advertisers and how we wanna engage in the brand advertisers and the performance advertisers and what this looks like. But, yeah, long story short is, you know, this is an opportunity to create a a brand for advertisers. That's what Geo is basically telling us is brand advertising is working in in AI. And we see a similar thing is that this is a top of funnel opportunity. People are switching their behavior, but it's slow. They're not buying things off of ChatGPT. They're buying things off of Amazon, but they're, like, doing their discovery and their research on ChatGPT. Right? And so that's what we see as well. And so, yeah, I think the main area that we are looking for from a distribution standpoint on demand is just people who you know, brands that are on the same page about that, right? Brands that realize that, you know, this is a consumer shift. We're actually able to get all this first party conversational data where the user's telling us what they Not necessarily like your blender example, what I'm looking for in getting a discount or whatever, but you might say, well, I really want my blender to be, like, from a sustainable company, right, like a company that, you know, aligns with my values as a human being. Well, where was that in the sale process before? Right? Now we actually can identify that, personalize the the data to them. And so, yeah, I think that's like all the opportunity, and so for to the extent that agencies are doing that type of work, and the GEO companies and their customers are doing that type of work, like, we wanna work with all the most innovative advertisers. So you mentioned you have advertisers that have budgets that are going unspent because the demand isn't there. Is that because there aren't enough publishers, which is probably one, obviously, one constraint? But then I I read in some of the materials that that I was reading through when I was researching you guys that you wanna be experts at intent. Right? And so right now, obviously, there's the one to one intent. I'm looking for a thing, an advertiser's got a thing, serve it to them. Yep. How deep are you going in, like, inferred intent Yeah. These days? Yeah. It's a good question. So one thing I'll answer, like, a non related question to that, which is that, yeah, part of the reason that we're not filling those advertiser budgets is because we don't fill random ads. We only wanna show an ad when it's highly relevant to the user. This super important for how we think about our product and our quality of the formats that we're building. Right? But the second piece is, yeah, how do we target? How do we figure out what intent looks like? And there are a million different signals that you can look at, and the kind of cop out answer, which is also just true from the technology and the way that it works is, you know, the ML model figures it out. Right? It's like, what is the combination of a hundred and fifty different signals that actually determines intent in this way, which is how it technically works, but then sort of what are the proxies that you can think about in terms of how that shows up? It shows up in a lot of different ways. Somebody might be saying, for example, that they're just feeling like really anxious today. Right? And they might be feeling anxious about financial struggles. Right? And maybe there's a product or an opportunity for, you know, like a student loan forgiveness program to kinda help them out. Right? And so it's kind of like, how can you extract that type of intent out of the conversations? And you can do some of that with the data itself. Right? Somebody says, you know, I really wish that there was a product that did x. Right? Because I'm, you know, within a conversation about a million other things. But there also might be scenarios in which the user is kind of just, I don't know, talking about their day or thinking through a problem or trying to figure out how to, like, converse with their boss, right, in which you can show a format that actually extracts intent from that conversation and says, hey, it seems like you're struggling with this. Would it it would it help if you had like a tool that helped you, like, sort of not have to think about your grammar so you can focus on, like, the the core of the product? Like, actually, that would be really helpful. Like, I'm not a great writer. I just need to think about the points, shove it into some AI, make it write well, and you're like, cool, Graham really does that. Yeah. You know, would that be helpful? I can give you a free trial. Yeah. And that's what an intent extraction might look like. So where's the company at today in terms of, like, size? You just got a twenty million dollar check. What are you gonna do with it? Like, where do you think, you know, how do you scale Yeah. From right now? Yeah. Hundred percent. And so I think we are gonna, you know, put a lot of that into the team. It's about half on the tactical side, half on more like the go to market side. Obviously, the go to market side will scale much more quickly Yeah. Get a lot more publishers on board, bring the advertisers in, make sure that we have sort of like a healthy marketplace. But the other part of this is that because we are bringing the advertisers in and bringing the publishers in together, if you think about how to serve the most relevant ad at any given moment, it sort of by definition requires that you have to have as many ads as possible. Right? The most relevant ad needs to be on the platform in order to serve it. Right? And so a big portion of kind of our strategy here is, you know, we're we're a subscale marketplace that has to grow in the same way that Uber did, DoorDash did, Airbnb did, and so there's a lot of subsidizing the market that has to happen as well. And this is basically the price that we pay for being able to focus really, really ******** on what the future looks like, and not trying to sort of maximize, like we were talking about just before, the revenue that we get today from like a bidding system or whatever allows us to focus on the best possible formats, the most relevant possible matching system, right, the right types of ad to query matches, and what that what makes sense there. Even things like what happens after the click, is there an after the click? Is the click the right way to measure success here? Right? And so the the money that we've raised here is, you know, really exists primarily to scale liquidity in the marketplace, and there are a lot of different ways we can do that. Because if you think about what the ad creative could be, it's almost infinite. It's infinite. Right? Not only could it be a something static, it could be dynamic, it could be its own l l it could be its own chat That's right. Potentially in there. It could take you to a place that has a whole personalized website created based on what they they were interested in. That's right. Like, there's infinite ways, and so that landing page or the click ID or whatever that you're right. That probably isn't the, you know, measurement of success Yeah. In this case. That's right. And so what we're doing right now is figuring out which of those things work, what come in, and does it matter that the user lands on a page that has context of what they just came from? Probably yes, right? And so we're basically figuring out is there an API that we can build to give the advertisers some portions of the signals that we found so that they can rebuild the experience and personalize it on their side. So doing all of those things takes time, it takes head count, you know, and it's we don't we're focused on that. We're focused on the best possible experience right now, and that's gonna take some time to build. So I've been on the marketing side for, what, oh, jeez, like twenty five years now, and and on the b to b front, account based marketing, digital account based marketing, when, like, Terminus and AdSense or Sixth Sense and and all these ABM platforms, you know, popped up, there was almost a a not infinite, but a, we trust you kind of budget going to marketers to be like marketers are like, look. I they're not telling me who the people are, but I can tell you that we're marketing to these companies, and we're and they they say that they're connecting to LiveRamp and giving me people who are c suite, but, like, I don't really know. And when they come and they click on an ad, I can't tell you who did it, but, like, I they sometimes tell me when the company came, and there was this whole, like, well, you know what? This is new. Let's see how it goes. And their budget was like, let's, you know, go figure it out. Then you start seeing that contract because it's like, wait a minute. This is becoming a bigger and bigger line item. How are we proving ROI? Do you see the question is, do you see that becoming a similar trend where at the beginning here is like, this is a new ad channel. We gotta try it. Let's kinda pray and pray here and figure out what happens. But then over time, you're gonna have to figure out that metric of what the what success looks like because the justification of those ad dollars, especially for the brands that are like, let's test it with a quarter of a million dollars Yeah. Yeah. Yeah. You know, it it it may not last forever Totally. In terms of A hundred percent, I think there's like one kind of caveat there in the AI world, which is that for sure the world is not gonna stop using AI, so it's always gonna be a thing. Right. Kinda like TV. Right? It's hard to measure TV. People spent ads right for fifteen years before we had CTV that could measure it. Right? And even an hour. Made it hard, right? That's so, I think we're more in that boat, right, which is said, you know, there will be evergreen brand dollars that will be available in this channel because we know the users are here, they're engaged, and all these things. But but also figuring out how to measure success in this channel is of critical importance to our company's survival. And what we've basically seen from ChatGPT and shutting down kind of the shopping stuff and even, like, their early ads results is people aren't buying stuff on AI right now. And so you kinda can't do a direct response measurement profile. You can't say, like, yes, one dollar in equals two dollars out. It's too hard to get that attribution right now, and maybe that will change over the next couple of years. But for sure, in the next, you know, call it two or three years, this is, you know, something that's gonna be a big question mark. It's like, is this working? How is this working? How do we know it's working? Like, what is this success metric? And I'll throw a couple of examples of stuff we thought about out there, which is, you know, can we measure brand lift within the chat platform itself? Right? And so we show like ten Coca Cola ads, do we see an uptick in Coca Cola mentions as a brand? Right? Yeah. Similar can be like a topic lift in the direction we wanna go to. Somebody can we prove that we can show that somebody is, you know, going deeper into the travel, you know, category as a result of us showing ads versus not? Right? And that would be only from the data you get from On first party. First party, from the advertise the the platform that's advertised. Right. Okay. Yeah. Or rather, the platform that is showing the ads Right. Or publishers. Right. Publishers. Yep. And so those are just a couple of examples, and there's a lot of other things that we can think about in terms of, well, if we show these interactive formats that go down and down and down, you know, we have one sort of experiment we call the rabbit hole format, right, which is just an infinitely generating app. Right? So if you engage with it, you know, via a button or a poll or, like, chat with it or whatever it is, instead of taking you off page, it will just take the the information that you gave it and then generate a new ad. Right? And you keep going and going and going. Right? And you're like, okay. Well, cool. Like, is successive like there? Is it the amount of time that the person spent in that ad format? Is it the number messages they went down? Is it the information that you were able to pull out of it? Are we able to enrich basically email retargeting of a user, right, to say like, awesome, they this user, you know, came and said like, I wanna look at Patagonia. I care about sustainability, repairing my clothes, and, you know, durability, and whatever. Right? And then you basically package that information and send it over to the advertiser, and that means that that cohort of people that they then are able to send emails to, or retarget on their websites, or whatever it is, are, you know, ten times more likely to convert. The beautiful part about you guys being an AI company is you probably because where my my mind goes, ****, you're gonna need some good account managers who can speak to these brands and, like, massage it a bit and be like, well, look, this is happening, this is happening. But couldn't you just do that with Like, case? Can't you just have the agents going and doing the research on this stuff and sending them reports? Yeah. You know, and have yeah. You got a couple enterprise account managers who are gonna hop on calls when needed, but, know, I don't know. Seems like a pretty good use case for, you know, slimming that team down. Totally. I mean, agents is a thing that we're investing pretty heavily in, and I think, you know, there's no replacement for the human to human relationship that account managers have. The question is, can we empower the account managers with basically infinite tooling? Right. Can they manage ten times as many clients? Or similarly, can you just empower the advertisers themselves with those tools? That's right. And and, you know, that way that they can get all the information that they need in a pretty adaptable way. And so we're thinking about this quite a bit in terms of, you know, how do we leverage basically, like, AI tools internally to make us and our especially our go to market motions sort of much more efficient, but also allow our cup our our customers, you know, advertisers and publishers both to have more access to, you know, dashboards and datas and things that they wanna look at. It's like, great. If somebody wants a certain cut of information, you know, we can't expose a SQL, you know, interface to them. But what we can do is, you know, create a couple of of, you know, post post response, you know, roadblocks, say, don't share this type of information, but, yeah, no, you can create and cut some different ways, and and that way the advertisers can get the reports that they're looking for. Alright. Fast forward twenty thirty, what does the world look like? Yeah. I mean, I think I alluded to this a little bit before, but we have a pretty strong belief that, you know, the Internet is basically gonna move to this agentic, or at least dynamic experience. And the reason that we believe this so strongly is that you you basically can just look at the engagement numbers and look at the usefulness of these things. Everybody's moving to AI for their search and discovery, but why is that the case? It's the case because it's just a better experience. Right? Like, Perplexity has twice the session time as Google. Why is that the case? Because it's a better experience. People are going deeper. They're spending three times the number of searches on Perplexity. Right? The same is the case on the entertainment side. I mean, looking at, like, character AI and janitor AI and some of these sort of, like, entertainment based chat experiences, and for what it's worth, you know, to clarify, those are not really people who are our customers. The average engagement per user per day is north of three hours, which is double what YouTube and TikTok is. Right? And so this is just such an engaging and such a useful experience. The reason that we're not seeing it everywhere already is in part because the technology still has to basically distribute and and sort of get good at all these different use cases, but also in part because it's really expensive to scale. And so we are here, the reason that we exist as a company is to get rid of that part. Right? Make it not a blocker to scale these great and super engaging and super useful experiences. You know, caution be a blocker to scaling those things, and that's why we exist. When you say, just so I can see if I'm I'm on the same page, When you say that the Internet will be a, you know, as I guess, user led Yeah. Experience. Right? Dynamic. Yeah. Does that mean that there there's no reason for me to go to a website. A website who who cares about going? In a way, the the interface of LLMs right now being mostly text almost feels like Microsoft dot or whatever that, you know, dot made the green and black screens. Right? You're like, oh my god. Look how antiquated that feels. Is it gonna be where in the future? I'm just asking it, whether that's voice or whatever, and it is creating a web experience for me that is curated to whatever I'm looking for. Yep. I mean, is that what you're imagining? It's probably not gonna be, you know, totally like that. I think, you know, we we say this a lot of times at Coat too, that we're bullish on screens. So we do think that people like the visual experience, and we do think that brands are really great at figuring out how to create that experience and building their stores, and we don't think websites are going away entirely. What we do think is that users will come to expect basically like a very excellent level of personalization, and also an excellent level of dynamicism within the website. Right? Cool. I can land on the website, and I can just tell the website why I'm here. And the best case scenario, they can do that without having to type. Right. So can there be some type of inferred this is why you're here? Can we bring that data in from maybe a chat agent that they were working with, or their personal agent that they, you know, have some information about that's willing to share with the website, you know, here's the experience they're looking for. So that's basically the way that we think about it is, you know, I don't think that websites are going away necessarily, but what I do think that will happen is, you know, a land a homepage from a website that is applicable to all people is likely not an experience that's gonna preserve itself for very long. I would expect all of those web pages to, you know, personalize to the end user and try to figure out more information about that. It's almost like a futuristic version of the UTM parameter coming in, and like, you know, whether that comes in through API or through whatever it's passing through, and you Or where a user just tells the platform. Right? They do, right, I'm here for this. Right? Great. Awesome. Let me help you with that. Right? But you're right. Each website today can't access what unless unless the previous website embedded something in a UTM parameter that passes through, they have no idea That's right. Where that person came from, you know? And so if there was a centralized place that was sending data through, that was allowing a website to personalize based on what they've what the other platform knows about That person, that could be really interesting. Right. Or that the user is opting into that experience. Right. If they have a personal agent that they're bringing that data with, and there are companies that are looking at this, and somebody should should build that protocol. Right? Like, we don't really want them to get into a bunch of, like, cross platform cookie tracking and stuff like that. We want the users to say, please take all this information that I just figured out with my agent over here and not make me have to rehash it out over here in this ad experience. Right? So that's that's the problem we're trying to solve. Yeah. Awesome. Alright, Nick. Thanks for coming by. We appreciate the time.