AdSense for the AI Era: How Ads Are Being Completely Reinvented

0:00 (logo whooshing) - Talking with Mike Choi, co-founder of Koa, they're building what everyone's calling AdSense for AI. Koa is the monetization layer for AI applications, and they just raised over $20 million from Theory Ventures. They're already processing over 170 million queries and seeing over a 5% click-through rate, which is four to five times what you see in traditional display. But here's what makes Mike's story especially interesting. He's a Korean national who came to the States as an exchange student in middle school, served in the South Korean military, then worked at Apple, then under Elon. And now he's building one of the most important

0:45 infrastructure companies of the AI era. We're going to talk about why AI ads are fundamentally different from search and social, why conversations aren't there yet, but will be soon, and what it actually looks like to build the advertising stack for a new computing paradigm. Mike, so excited to have you on the podcast today. Thank you so much for being here. You were a Korean exchange student in Kansas, jailbreaking iPods, and now you're building AdSense for AI. You have one of the most interesting stories we've ever had on the show, and I'd love for you to share with the audience. - Yeah, thanks, Anthony. Yeah, it's a very long story,

1:26 and I'll try to make it as pleasing or interesting to hear. But yeah, I was born in Korea. My dad was in the Air Force, and my mom was a stewardess for Korean Air. And I guess they both met in the skies. But my dad, part of his officer's training, had to go to the States, and this was around when I was in the third grade. And so I had the opportunity to kind of follow him along with side my mom to go to Arizona, or Tucson, Arizona, to be specific. And yeah, I just kind of fell in love with the States. I kind of, you know, any time I tell this story, I kind of mentioned the moment I had at my friend's house

2:15 when he made me like a Nor's instant rice, you know, broccoli and cheddar, extremely processed meal after school. And I can like still remember how that tastes in my brain and kind of like change the chemicals up there, maybe. That kind of made me go, wow, this is great. I would love to stay in the States. So after a year in Tucson, I went back to Korea, and you know, just things just didn't feel the same. You know, it's kind of like going back to the same Korean academic system and all those kinds of things felt a bit like I was missing something back in the States. And so I asked my parents to send me back

2:57 to the States as a foreign exchange student, essentially. And then I found my way to Kansas, where I stayed with essentially a host family where the mom was white and the dad was black. And you know, the kid that they had was in my same grade, was a mixed kid. And then me, the Asian kid, ran around as well. And so it was a very international family. And I learned tons there being surrounded by extremely smart people and, you know, parents who were very generous and nice to me, just like a kid from really nowhere in middle of Kansas. And, you know, I was gifted an iPod for Christmas and started downloading Cydia, you know, the PDF jailbreaking websites

3:44 to kind of really customize my iPod and became an Apple fanboy, essentially, right? And that basically kicked off my love for consumer apps, building these apps and software that brings joy and delight to people all the way, you know, from like designing it, building it, shipping it. That just brought me an immense amount of joy. And so that's, I would say, where I started my love for apps and that kind of turned out well for my professional career because, you know, I went to my dream job at Apple and then essentially eventually Twitter. And even to Koa, which is kind of, you know, inspired by my high school moment where I was making apps

4:33 and had a hard time monetizing it. At the time, AdMob was kind of like the de facto solution if you wanted to monetize your application. And at the time, the state of the art was having a fixed banner or an image ad on the bottom of your tab bar. And anytime a user clicks on it, you get paid for it essentially, right? But now it's 2025, 2026, and there's ad blindness to those banner ads. And they're not really something consumers like. And app developers don't like those banners as well because it makes your app feel cheap and it doesn't respect the amount of craft and effort that went into creating those apps. And so at Koa, what we're essentially doing

5:18 is let's reimagine what ads look like, especially when there is this entirely new surface area that is changing what feels like every hour of the day. And so, and yeah, we're well in the journey and we have some thesis and products that we think are the right tools for the job. - Such an incredible story and love that your background has really culminated into what you're working on now. And I think what's interesting is every major paradigm shift in, I'll call it media or information or content has come with some major re-imagination of advertising, but it never goes away. So if you think radio to TV, to web, to social,

6:09 there's just whole new infrastructures built around that new paradigm. So you're calling this AdSense for AI. How is AI going to make things different and what's the answer to it? - Yeah, I would say a lot. I think when we first started this business, the 50,000 feet vision that we had was, hey, every single ad that you and I see on the internet is practically the same. It's been created by a creative agency, checked off by a person at a company that says, yeah, that does speak our brand, that does meet our guidelines and is circulated around the world. But what if everyone is served an ad that is just tailored made for them specifically, right?

7:02 That is marketing holy grail, right? And so AI is able to do that at scale. And we as a company are trying to figure out, hey, what does that look like? Is it as simple as just saying, hey, Mike, would you like insert brand name here's shoes? Or is it dynamic interfaces where the UI is generated and the ad is generated and matched perfectly to fit the foreign factor of the AI application, right? And so I think there are a lot of creative ways to do it. We started with the text generation where every ad was generated on the fly, right? To kind of match the context of the conversation. And now we're taking it even one step further

7:44 by saying that, why stop at text? Let's go one level up, which is the assets. And what's one level up above that? Well, it's the actual ad. How does the ad look and feel? What would be a good example of a static old world advertisement and then advertising with Koa? Where have you seen some of the biggest successes with your customers? - Yeah, great question. Yeah, I'll give a static very simple example for what a quote unquote traditional ad might look like, right?

8:24 Let's start from like the real world, right? You're in a New York subway, you see an ad. That's the same ad that everyone sees when they are writing that carriage, right? But let's say in Koa's real life physical world, it's a version of the product, every single Visa ad or every single insert brand ad here, the text, the content, the product that's being sold is tailored just for Anthony or just for Mike, right? It knows that I have preferences for certain products and it knows that I have, that I'm not really a fan of other category of products, right? And so it's able to kind of magically know and sense what the best ad would be that not,

9:15 that doesn't feel like a encroachment or anything, but more so a value ad, right? It's the moments where the ad shows up and I go, "Yeah, I didn't know I needed that, "but yeah, I need that right now," right? Those are the moments that we're trying to unlock and it's super hard to do so because if you're slightly off, then it feels creepy, right? But if it's just right, there is this like moments where it's, "Wow, yeah, thank you for showing me that ad," right? And so those are the moments that we're trying to unlock via the dynamic generation. And of course, there's some ad tech 101 stuff that we also have to nail down such as like the,

9:53 how do I match a user's intent to thousands or millions of ads that are living on a database, right? So it's a culmination of various factors that makes it feel dynamic and alive and very well contextually relevant. - Do you think this is gonna open up new distribution channels of advertising? I mean, social is still dominating the advertising world right now. Do you think we're gonna see more creativity in the chat harness over the models or is there some other way your envisioning consumers will interact with advertising? - Yeah, I think, yeah, there's two questions there, right? I guess the first bit is, you know, other than social,

10:39 other than Instagram and the typical platforms, yes, I do think there will just be an explosion of those platforms mainly because in the past, I'm sure you have that one person in your life that's like, hey, man, I got a really great idea for an app, let's build it. Especially if you're an iOS or an Android engineer who knows, you know, your friends know that you can build an app, you've shipped an app before. They wanna convince G-course you to say, let's build that recipe app, right? No one's done that before, right? But now they can actually do it. They don't need an iOS engineer in their life. They can just go to a vibe coding website,

11:15 you know, type up a prompt and then make an app and ship it, right? And so what that will lead to is two things, right? One is an explosion of apps, but secondly, it means that sure, there will be an explosion, but you know, there will be bad apps, right? Another person, that would be one person that's trying to make yet another grocery store or a recipe app that doesn't need to exist, right? But there will also be apps that had to exist or just adds value to the world, but just couldn't have been made before because the dev costs were too high, right? So very niche communal apps that just serve a lot of purpose

11:56 and value to very niche groups will definitely pop up. And yeah, and you know, and Koa is ideally there to kind of serve those apps monetize and kind of sustains themselves, right? Because the inference costs nowadays are, you know, it's non-zero, whereas in the past in 2010, your server costs were probably the only amount of bills that you had to pay to maintain an application. So yes, I think I'm bullish on the fact that there will be multiple platforms that will show ads to frankly, sustain themselves as a business. - Makes a ton of sense. And something we're seeing a ton, at LeanScale, we run GTM Ops for mostly B2B SaaS companies.

12:46 And the past couple of years, we've just seen an absolute dramatic increase in volume of outreach and people trying to get in front of other people. And now that you can automate so much of this, the noise has just, I'm curious, are we gonna see a similar thing with advertising where now you can just create an incredible amount of volume and how are people gonna stand out and stand out through the noise? - I think that's a good question. I think it really kind of depends on the platform or the ad platform. The advertiser is putting ads on because at the end of the day, it's really the ad platform's choice to show which ad, right?

13:34 And so I will tell you our beliefs and what we optimize for. And I don't, I'm sure Meta has their own principles in terms of which ads to show when. And that's a whole, it can't be covered in the next 20 minutes, probably. But for us at the end of the day, what we believe is that we wanna build for the end user. The user that is actually seeing and experiencing our ads. Those are the people that we are optimizing for and trying to add value for first. The reason we're doing that is because, hey, if Joe from, or Sarah from Arkansas is experiencing and is touching and is interacting with our ads and are gaining value, actually going,

14:25 oh yeah, that was a really cool product or that was a really cool experience. Then their engagement with the app that the publisher built will increase. And from engaging with those ads, the advertisers will also be happy because they're there to optimize or to accomplish a task that that format is designed to help them accomplish, right? And so, and finally, I always, one more thing is, we just believe that the surface area, the real estate, the publishers have granted us is sacred almost, right? It's the apps that these developers are building are kind of their little babies right there. They're putting a lot of craft,

15:05 they're putting a lot of care into those apps. And we think the ads deserve the same amount of care and attention. And so kind of putting the emphasis not only on the UI, but also asking the question of, is this valuable? Is this adding value to the end user? And of course, looking at the metrics to make sure ad ID X is doing that is our newest star. - Well, I think in order for anyone to keep up the hyper personalization component is going to be key. It is showing in your rates right now, 5% click through conversion rate, which is incredible three to five X typical benchmarks. Do you see that continuing to stay at that level?

15:53 Do you think it'll commoditize at a certain point and come back down when more people are personalizing their ads or where do you see the trend going? - Yeah, I think if I were to tell you exactly where it goes, I would be a magician and exactly know how the business will Japan out for the next five, 10 years. But I can tell you what I think, right? I think CTR is an interesting measurement because sure the user can click and the number of clicks with the click through rate can be high. But if the user isn't converting at the end, then the CTR really isn't the best way to measure success. Now that depends on what the objective of the advertiser is, right?

16:39 But we actually think at the bottom of it is, if we are doing our job correctly of serving the most relevant ads or ads that will add user value, keep saying that again and again, the CTR engagement rate or conversion rate or whatever the advertiser is optimizing for should stay the same, right? There's no reason for it to decrease. If it does decrease, we're not doing a good job, right? If it increases, well, we're knocking out of the park, right, and so, and along with some of the products that we're building right now, it's ad blindness is not going to be a thing because of the generative nature of the UI, right?

17:21 So we think a lot of the patterns or know-hows that were learned over the last couple, last decade or two of ads will be completely irrelevant, kind of like how most PM books or most engineering books that were written before this year are practically just out of date. They're just no longer irrelevant, right? And so we think the same will happen to ads, UX and, well, yeah, behavior. - Yeah, and it does feel like things are changing at such a fast rate. It's pretty tough to keep up. I'm curious, maybe going to the origin of COVID, what was the hardest thing about getting it off the ground and bringing this to market? - There were a couple,

18:10 but I would say the main thing that was hard was we, from the get-go, we kind of felt like the idea was obvious. It was an obvious one. We had the idea, I would say even like a year before when we actually started the business. And at the time it was around when GPT 3.5 was out. And it felt too much like a slot where people weren't productionizing elegant features yet because they felt like it was too early, right? And so we kind of shelved the idea, put it into our backlog, hoping that the right time will come. And after a year, we thought the timing was right. And so initially when the idea first came to be, the problem was, are there even apps out there

19:00 that are using LLMs as like the main feature, right? Sure, there could be a chatbot here, chatbot there, but are there enough of those apps out there that justifies a billion dollar or a trillion dollar company to exist, right? So that was big question number one. And number two is I would say, well, that the first problem I just mentioned still exists today, I would say. But the second question is, what is user behavior in AI like? And what is the perfect ad in AI? We still don't know, frankly, right? And I'm not sure if we'll ever know, and that might be a controversial statement, but we are the ones that are trying to kind of define the standard in AI ads

19:52 to be able to say, yeah, this is what, this is what state of the art looks like. This is what working looks like, right? And some of it I think is stuff that we can control via our technology, improving our engineering, improving our design. But then there's also a different bit where it's just pure consumer behavior, right? Shift in consumer behavior. And that just takes time. And I'll give you an example, right? If I want to buy this microphone, right? I wouldn't go to Chachipiti to buy it, right? I would simply go to, well, sure.com, or I would go to Google and type sure. And then I would browse through all the platforms

20:33 and providers that are selling this microphone, buy the cheapest one, use my favorite credit card to get points, and then go from there, right? And so there are, that's the kind of consumer psychological gap that we still have to cross. But meanwhile, there's still tons of other objectives and goals that AI ads can help advertisers accomplish. And so we're looking at those opportunities and winning in those areas. - Is there a certain inflection point or signal that you had that gave you the confidence that it's moving in that direction enough? - Yeah. - To make you jump in? - Yeah, well, I think, you know, when we started,

21:16 we had the thesis of, yeah, just, you know, if the user is asking about shoes, show a shoe ad and then they'll buy it, right? That was the very simple naive hypothesis. And it turns out people don't do that, right? Because of various reasons, which one of them being like the psychological gap that they have sought to cross. But the good thing, the side effects of that was that we developed technology to be able to track and know when the user does convert at the end of the pipe, right? And so we always had kind of visibility into the actual sale or the actual conversion that was happening down funnel, right?

21:56 And so I would say like starting a couple of weeks ago, actually, oh, like a month ago, that number started to go up, right? And so we're like, oh, something's happening, right? We sold X, we sold Y and the numbers are growing up. There must be something happening, right? So either we did something with our ads or the serving technology or the advertisers that were on the platform or people are starting to feel more comfortable or they're starting to feel like, yeah, let's try it. And then kind of at the end of the day convert, right? And so I think the decision to kind of invest in that analytics component of the business early on has been very fruitful.

22:39 - That's fantastic. I'm curious, you used the sure example. I definitely had to go to an LOM to figure out what are the best mics? What should I actually get in it? - Yeah, yeah, yeah. - But I didn't buy through there, so I got to consumer behavior. I've heard anecdotes that the price point dictates how much research they may do and the likelihood of maybe going to an LOM to do some of that research. Do you have any data on which products are most successful and get the best ROI from leverage ads? - Yeah, yeah. I think that's 100% right. The more expensive or more complex the product is, the more research the user wants to do

23:27 before actually purchasing that item, right? So what we've seen from the product category perspective that kind of fits into that are taxes, student loans. Anything financial related or credit cards, right? Has been very, much so in that category. And what we've seen recently because of seasonality that works is taxes, right? You should see our numbers on tax day of people clicking on tax ads was incredible, right? Because some human behavior just doesn't change. And it's very consistent. People don't file their taxes until the last moment, right? And so I would say that's one of the things we've seen

24:16 that really works for high consideration products, yeah. - What's next for you, the team at Koa? What does the next frontier look like for you and the team? - Yeah, I was always asking that question, but right now we've thought about this for a bit and we're on a path to really build the dynamic UI component of our business. And that is, like I said earlier, we started with dynamic text, but we're allowing or we're building the technology for any market or any advertiser to build an ad format that really uniquely represents their brand, right? And so you're not confined to a single JPEG. You're not confined to a 50 character limit.

25:07 You can come to Koa, use our dashboard and prompt away as a marketing engineer, right? And say, I would like this, this, and this in the ad format or, you know, and here are some components of our brand that uniquely speaks our brand and craft a unique sponsored experience just for your brand, right? That is something that will be delivered to the end user and they can engage with it. They can click out, all those kinds of things. But I mean, the cool thing that we really want to enable is yeah, static UI, just showing a JPEG is very early 2000s, right? LLMs are moving so quickly. Let's build towards the future and what it can do instead of going,

25:57 oh yeah, what do we do in the past? Like, let's just do that first 'cause we know it works, right? Because I think building for the future is more exciting and I think there's tons more creative things that marketers, really talented designers could do within our platform that we might not even be able to imagine ourselves, right? So we are building the tools, the Lego bricks that then they can use on their own to kind of craft these beautiful experiences. - Such an exciting time. I'm curious. I have a lot of people, especially at LeanScale, we're heavy on building AI workflows and automations for our customers. So we're definitely like deep in the space.

26:40 How do you stay connected? Where do you go for your information? Where do you go for inspiration of what you need to be building next? - Yeah, well, for tech related news, since having worked at Twitter, I still have X or Twitter on my phone. And yeah, my coworker was like, where do you get all this? Like, how do you stay so connected? I was like, oh, I doom scroll from time to time, right? Should probably stop doing that here. But for the products and those kinds of things, I've recently been reading and purchased the Great American Ads. It's like a coffee table book and there's one book per decade, right? So the Great American Ads of the '40s

27:24 looks very different from the Great American Ads of the 2000s, right? And so there's the Got Milk campaign as one of the examples in the 2000s book, whereas in the '50s, there's like a cigarette ad or an ad that probably wouldn't fly today, right? And so kind of looking at those and being like, wow, these are so beautiful. These ads were pieces of art in the past, right? And now it has certain negative connotations, especially online. Why can't we bring back the days where we go and wow, that is such a beautiful ad, you know, there's a full pager ad that a person might take out on the New York Times and go, let's rebuild that.

28:04 Let's make sure that whatever artifact we produce at Koa kind of makes people go, wow, that is a cool ad. That is also one of our north stars. - I love that. And somebody's got to do the doom scrolling and fill everybody in, so I appreciate that for us. - Pleasure to be a service, yeah. - Mike, this is a really, really exciting conversation, especially timely for what a lot of our customers, people who listen to the podcast are trying to work towards. And I think in order to maintain your level of competitiveness, you're gonna have to personalize, build meaningful messaging, imagery, copy, that speak to the person who are looking at an advertisement.

28:51 And I really appreciate the work that you're doing. And it's reflected so much in your personal story, all of your experiences, kind of culminating into jumping into a life of entrepreneurship, birthing something to life, and really being at the forefront of is one of the most exciting times in human history. So I just appreciate everything you've shared, what you're building. And for anyone who's listening, what's a good next step? Where should they go take a look at? Where would you guide them? - Yeah, if you want to talk about what it's like to start a company, or any questions about what is a startup like,

29:36 I'm available on Twitter, the handle is guard_if, and some of the Swift programmers out there will kind of understand what that means. And if you're curious about what ads and AI looks like, and what the frontier feels like, go visit our website, coallabs.com. And we have a new brand, and a visual guideline dropping soon, too. And it's gonna be really fresh, and I think the folks will like it. And so stay tuned for that as well. - Mike, thank you so much, and can't wait to see what you all build next. - Thank you so much, Anthony, I appreciate you.