#63 Growth in 2026 - The REAL IMPACT of AI on Work, Startups, and Creators ft. Casey Winters

Mentally, like nothing feels really hard to build or replicate in nature of AI. What a lot of startups are doing right now is they're trying to build acquisition before retention. So there's this kind of rollercoaster of, you get a bunch of people trying it. And then it goes like, yeah. Is there any part in this that you feel, let's say teams are over indexing, which is like, this is like a lot of hype at this point in time, versus under indexing where? A lot of people are saying like, look, the new distribution channel is going to be, you know, AEO, right? Like stringent optimization. And I think that's largely a trap for, for most companies. Then like the LLMs have openly stated that they want to build agents that do just that for every vertical, right? The incumbents are smarter about, you know, doing that well. And they're just doing it a lot faster this cycle than did last cycle. Yeah. And now it feels like you have to do it every year and things just change too quickly. Right now we're, we're early enough in the cycle that people are growing revenue and users really healthily without that. But I think the rug's going to get pulled out from them, you know, eventually let's call it by 2030 at the latest. It's going to tank. That's true. Hey everyone. Welcome to the Founders Foyer with me, Aishwarya. This foyer is full of conversations. The space where creators, founders, and builders look for all the support and concepts to grow their ideas into products. In a world where generic co-pilots can answer any question with confident bland paragraph, the real bottleneck isn't information. It's access to the highly skilled operator judgment that actually moves a business forward. Casey Winters is rebuilding this layer. As the co-founder and CEO of SuperMe and one of the most trusted growth advisors in the Valley, he's turning private work, such as docs, notes, talks, and real decisions into AI profiles that you can actually talk to. So founders can tap the instincts of proven operators instead of AI slog. In this episode, we'll dig into how his thinking on distribution and modes has evolved in the AI era, why he believes AI is a technology shift more than a distribution shift, and how SuperMe is trying to protect human expertise in a world full of autocomplete. And this is Casey's second time of being on the show. Casey, thank you so much for being here. It's always a pleasure to chat with you. Yeah. Thanks for having me. Awesome. So growth has always been an on-demand job, Casey, and with AI, I'm only seeing that there are more roles that are coming up in every company pertaining to growth. Has growth as a function evolved from, let's say, being there for SaaS-first companies to now AI-first companies? Do you see anything particularly that's standing out? Yeah. So, you know, the concept of like growth as like separate or different from marketing started with consumer internet because with software, you could change it. And it turned out a lot of the biggest drivers of, you know, growing users or growing paying customers with software was like a change to the product that made it more accessible, you know, whether that was to search engines or made it more viral or retained better, monetized better, things like that. Yeah. And so that concept of, you know, efforts in the product to drive growth made its way into B2B SaaS over the last decade. And I think in both kind of consumer internet and in SaaS, before AI, we typically talked about companies having like distinct phases. There is a zero to one phase, you know, get to product market fit. Yep. And then post product market fit, you just started emphasizing a lot of efforts on growth, right? Let me go get more customers. Let me make it easier for them to find value, make it easier to monetize, all that kind of stuff. And then way later, if a consumer business, you might eventually, you know, invest in some new products and, you know, maybe three to five years later in B2B SaaS, you would invest in new products and sort of expand a suite. I think the change with AI is that due to both rapid underlying model improvements, as well as the cost of product development going down with AI coding tools, it feels like you have to do a lot more product reinvention along the way and net new product development, meaning like new features, new product value being developed to drive that growth, rather than just being like, I have sort of a stable core product and I just try to get more people using it, more people paying for it. So there's much more of a blend between what we think of as, you know, growth work and, you know, new product development work, new product reinvention sort of work. I mean, we used to not think about product reinvention until companies were like 15 years old and now it feels like you have to do it every year and things just change too quickly. Yeah, true. And it's almost like not even every year, I guess with AI, it's every couple of weeks, there's a new... Yeah, it's probably generous. Exactly. And this is a very interesting point too, because when you started out as what growth felt like for consumer internet and SaaS, it almost feels like, oh, we're talking about acquiring people, expanding the business and then keeping it sort of like going very steadily. But I think with AI, the newness as in the product roadmap also gets to be very much a part of the growth engine. So it's not just that you want to grow in terms of the users and broadly, but also grow in terms of the product coverage and what it gives. That's interesting. And is there any part in this that you feel, let's say teams are over-indexing, which is like, oh, this is like a lot of hype at this point in time versus under-indexing where it is still the most important work, but it's unsexy and like a lot of compounding has to happen. Has that quotient changed in an AI first business? Well, I think what you're seeing with startups, at least, is a reaction to the fact that while there's been a lot of innovation in what they can build, there hasn't been a lot of innovation in where they can find cheap customer acquisition. So what's strange right now is, you know, I define product market fit as satisfaction that creates sustainable growth. And what a lot of startups are doing right now is they're trying to build acquisition before retention. You know, that might be like, you know, on TikTok or something like that. And retention is usually the main indicator that you have, you know, satisfaction. And so what that means is like you're doing a bunch of effort to drive hype that doesn't create that sustainable growth because the people just don't stick around. So it seems like wrong on both fronts. You're not creating satisfaction and therefore you're not creating sustainable growth. So it's this kind of rollercoaster of you get a bunch of people trying it and then it goes like straight down. And I think part of the reason for that is the venture capital industry sometimes rewards that hype with more capital and that gives you more shots on goal. But again, a very small part of your role as a founder is to play the investor game versus the customer and user game. And I don't think people are think they're over indexing on the venture capital game. I just think people are starved to find ways to get attention for their products and they're doing it in pretty inefficient ways, bringing in either the wrong customer or bring them to the wrong product because the way they can drive attention doesn't match the actual product value that they can create. So there's still a lot of mismatch there. I'm seeing versus like, OK, well, how strong is our attention? You know, how well can we monetize retention and how can we reinvest, you know, satisfied users and the money we make from them into, you know, good ways to acquire, you know, more customers, more users, things like that. So I think there is a bit of a mismatch. And I think that will evolve over time as, you know, this next generation of founders gets smarter about acquisition. Right. Yeah, that's so true. And does Velocity actually have a reason to play in this? Because just a minute back, we were talking about how release cycles, all of these AI advances are happening with speed. So I'm also assuming that with that speed, let's say there is a certain, you know, model launch or there's a certain feature or a module launch this week. And then we don't have enough time to even let alone have a release cycle. To even like, let alone have people see it, let alone have people use it. And then two weeks later, there's something else. So does speed, as much as it helps in other things, actually play a sort of anti-role here, you think? Yeah, I think the environment we live in right now in tech is creating the impression that it is impossible to keep up with the industry. So there are people who try to take the lead on like, I'm going to explain to you what matters. You know, the game has changed sort of post on, you know, LinkedIn and X type stuff. And I'm sure that's good for content creators, but that's not necessarily good for us in terms of building products. So I think it comes back to, for most, not all companies, you know, the Steve Jobs logic of like, you have to start with the customer problem and then work backwards with the technology. As long as you know what you're focused on, which problems you're trying to solve, and then you're looking into what are these new tools coming out that I could potentially add to my toolbox to solve those problems better, then you're going to be in like a fine position. If you think you have to rethink everything with every new open AI or Anthropic or, you know, Gemini announcement, then you're going to just be like consistently distracted from what your actual customer is telling you. And that's where things can get off the rails, right? So I think, you know, we're learning obviously a lot, you know, building super me, but we know like what we're trying to build, we know who we're trying to build for. And then when there's a technical problem related to that, then we go look in the market of like, has anything come out recently that's going to help us? And if not, like how far can we go in developing the ultimate solution we want? And when might there be another unlock technologically that helps us get like closer to the best possible outcome we're looking for? Right. So in some sense, it's also like keeping away the noise rather like not constantly falling into, hey, what's this hype cycle all about? And what are the other 10 tools that do the same thing versus like really going back to the fundamentals of problem solving? Yeah, that makes sense. Fair enough. And also one shift in the same theme that we're talking about. So because we're going back to the problem solving and really trying to understand the core purpose, so does this also mean growth strategy is no longer about just the release cycles as in founders cannot think that, oh, like now this week we're adding AI to the cycle. So it means shipping AI features is equal to building a distribution mode. Or that's my growth strategy. What do you see as wrong or right in that attitude? Well, I think, so first off, I think companies are getting better at this. So, you know, let's call it a year and a half ago, a lot of people were just like, you know, slapping AI onto the side of something and being like, well, now I'm an AI company and I have this AI product, you should pay more for it. And people like, A, I'm not using it. B, I'm definitely not paying more for it. And so now people have moved on to the kind of the next evolution of the cycle, which is like, no, I've actually built something we couldn't build before. That's AI native, that's actually helpful. And the problem with that is not that people aren't using it. In many cases, people are using it and it's growing quite rapidly. It's that, well, everyone else can do that too. So if there's no differentiation between, you know, this sort of magic trick that AI enables, then long term, there isn't going to be, you know, any distribution advantage or any like revenue advantage that that feature will just become free with dozens of startups and incumbents alike having built something very similar. So then you're back to kind of the more traditional phase of like, well, how is what I'm building creating a compounding advantage over time? And will I have enough time to create that compounding advantage before it's a commoditized product? And that's, you know, network effects or deep customer integration. And I think the problem we face right now is people are writing a playbook with this new technology as if it's the last sort of technological evolution, which is SaaS. And a lot of SaaS wasn't built on network effects or brand or even like switching costs. It was built on like, yeah, the thing I built is hard to build and hard to replicate. And fundamentally, like nothing feels really hard to build or replicate in the age of AI. So even if you've got, you know, a lead and you built something really cool, a company that is definitely much larger than you can build basically the same thing in a few weeks. Yeah, that's true. And that creates a lot of challenges for startups. So I think there needs to be more thinking about long term defensibility and differentiation. Right now, we're early enough in the cycle that people are growing revenue and users really helpfully without that. But I think, you know, the rug's gonna get pulled out from them, you know, eventually, let's call it by 2030 at the latest. And if you haven't like built something more defensible by then, that revenue growth is just going to go away really quickly. Yeah, it's going to tank. That's true. And in fact, as we're talking about differentiation, I'd love to get your take on. So whatever we're talking about applies to two layers, right? One is the application layer that we're talking about, which is the super me, there's like a bunch of other app founders who are building on top of these models. And the other layer is actually the core model layer, which is what OpenAI, Anthropic, Perplexity, a lot of these actual model companies, frontier companies are working on. So does the same logic apply to both of these, like both app founders and these model founders? Because when we talk about differentiation, what does differentiation really look like? Let's say for even these core companies where everybody's now competing to have a coding agent, everybody's now like launching left, right, center image model. So what's differentiation really like in this setup? Has it even changed in its meaning from what it is for an app cycle versus a model cycle? I think when these models came out, we thought they were so hard to build and required so much CapEx that they had the technological differentiation. And that was going to be a sustainable form of defensibility. And I think that has just not proven to be true in that, at least if you look at, there are enough companies who have put in the CapEx work, who have put in the model work, who have the right AI researchers to build largely competitive and very swappable products. And that's Gemini and OpenAI and Anthropic. Maybe you could throw Grok in there. Right. And so those companies, there are some ways in which the way they build product has to be fundamentally different. Right. So as an application layer company that's building a network effect oriented product, while we obviously sit on top of those companies and those are tools to help us build value for our users and ultimately our differentiation is going to be the network we build. And that's going to be our defensibility. But for them, I think what they've seen, which is a little bit surprising, is that, okay, if I come out with an innovation on image models or on web search or something like that, it's three to six months before it's available in all the other model provider companies. So they're like, okay, well, our tech isn't differentiated. But I think there is an element that's different if you're working like as a PM in those companies, which is, as I mentioned, kind of how Steve Jobs taught us all 30 years ago of like, start with customer problem, work backwards to the technology. That is not how things work at the model provider companies. Yeah. It's AI researcher hands you magic beans. They need to figure out what those magic beans are most interesting to use for and commercialize that. And so that's not what we do, obviously at Superme, but that is an important part of the role at, say, OpenAI or Anthropic, things like that. Right. But you're in a position at the model provider companies where you're like, okay, I'm growing really quickly, but I might need to think about defensibility just as much as my customers. So if you take a look at what's happening at OpenAI, I don't have any insider information there or anything, but they're launching these tools intended to give them shots on goal-led network effects. Sora is not released an image model, it's built a social network with the video model underneath, right? Yeah. And you could think of Pulse is trying to layer in more personalization to build data network effects, their investments in memory. So they're trying to think about like, how do I take all these users I got virally and lock them in so that if Anthropic comes out with a better model tomorrow, I don't lose them. And I think there's a good chance that that effort will work. It's just a shift in strategy. The other part that you mentioned is around how they all seem to be playing in all the spaces. And as a fundamental startup guy at heart, that's where I've spent most of my career, a large part of strategy is saying no, like saying no to a bunch of good ideas so you can focus on the best, most differentiated one for your particular business. And what happened last generation is certainly what's happening this generation is companies that get overfunded feel like they can say yes to everything. So if you're open AI and you have this big consumer lead with chat2BT, the Gemini is coming for, but you also have some of the best researchers on the planet and Anthropic's better at coding, you're like, wait, why should I give that market up? Right? Or enterprise, like, hey, I should also have chat2BT winning enterprise. So as a startup guy, the way it works in my brain would be like, well, Gemini is going to win developer tools, Anthropic's going to win coding, and chat2BT is going to win consumer. Why don't you all just focus on the areas, each of which is worth multiple trillion dollars and just go fortify your winning strategy there. And that's not what all three of them are doing. All three of them are trying to play in all the spaces. And I think what you saw last generation is what you'll see again, which is if there's any sort of crunch in funding or economic cycle, they'll start to make harder decisions strategically on where they play. Right? And they will make probably the most obvious decisions they could be making now, but they're sort of like, well, I got infinite capital, like why limit myself? Because I do have a 20% chance of coming back and winning the coding market. And that's worth an extra $4 trillion in market cap. And then, yeah, I'm going to take that chance. I personally don't like that way of playing. I like to be incredibly focused in what we do. But I can understand why it feels hard to say no to things like that when you have this excess talent, when you have excess compute, like all of those sorts of things that allow you to play that game. And if these aren't winner take all markets, and you're each cutting them by a third or a fourth, then that's another reason you feel like you need to play that game, generate this type of return based on the capital you've raised. But this is something that's rapidly evolving at the model companies. I do think they'll get better at strategy. You have to remember, most of these are started as research projects, not as traditional businesses. So they're like learning this stuff. You can't expect them to make like every perfect decision every time. Yeah, no, that's absolutely true. And a lot of this is like at an exploration stage. And I think the market is also now very comfortable accepting these experiments from even leading providers. I was just telling a friend this morning that a decade ago, people were just so not okay with Google doing these experiments. And they always used to have questions around, oh my God, like, why is Google naming the product so differently? But look at Nano Banana today. Everybody's like going nuts over it. And then people are like constantly exploring it and embracing that particular experiment from Google. And one of those experiments prior to that was Waymo, right? Which is just such an amazing product experience. It's growing incredibly quickly. Exactly. Very true. And it has been happening since like acquisitions of either like existing other products or newer experiments from within. But I think the mindset to kind of embrace that and play along with it is definitely come through. So I totally second with that concept of yours. And you're right. I think the free money is definitely making it making everybody want to compete. And I also maybe think hearing from you that is it because all of them are currently focused on the discovery engine part of it? Because when it comes to the retention part, which you just told a couple of minutes ago, then it only makes sense to go very niche into what really the particular brand would do for sure, which in case of Google and the rest is going to be more. uh, Google and the rest is going to be more around developer tools and with Anthropic, it's a different thing, but I guess at this point in time, they're also like way too much focused on, Oh, like I want more people to discover. I want more people to like start using this. So maybe it's also because of that. Yeah. So I think that makes sense for Google because they're trying to protect, you know, their current leadership position as a search engine and driving discovery across the internet. And Chattopadhyay is the first legitimate threat to that in 15 years. Uh, right. Um, that makes less sense for Anthropic, uh, to be investing in to me, given the lead that they have in coding and how powerful that is as a use case. Uh, but yes, that's like a defensive, that's like a big assault on their core business that they needed to react to. And, you know, up until recently, they, they, even though they invented a lot of this stuff that felt like pretty impossible for them to react to Chattopadhyay, they finally figured some things out internally and how to commercialize their tech, which has always been amazing. Um, and that's like, now I'm going to be very interesting to watch between those two companies and how they do get out for consumer mindshare. Um, and yeah, I don't, I won't pretend to be able to predict that. No, I think the best part is to kind of let it, uh, you know, roll it out and then, uh, and then sort of like see it in, um, action, but yeah, that's, that's so true. And I don't know if this is also a part of the growth, uh, uh, engine part of it, but then what I also noticed in the recent times is, um, the focus on a lot of dogfooding when it comes to models and literally having the entire company post about the outcomes of it. Like we saw how every second person in Google spoke about like playing around with the AI studio or playing around with these image models, like nano banana, and then every, um, anthropic executive talking about how this is being useful. And same thing goes with charge APD when they launched, let's say pulse or a ton of other features. So what, I mean, well, this happened in SAS as well, but not as popular as how it's happening now where people are literally dogfooding their own stuff and then using that as sort of the social mode to go and make their employees talk about it. So I don't know if that also adds to the growth engine angle of it. Well, I think there's two components. One component is, and I, and I think, uh, opening has done a really good job of this is that, um, if the biggest growth impact is going to come not from, let's say, um, optimizing my onboarding flow or getting more people to share, but by inventing a use case with a new model change that can go viral and get me a hundred million extra users, which is, you know, the, um, the studio Ghibli example, you know, their image things, then part of the job as, you know, uh, in, in, um, an employee of those companies is like, you need to play with the product to figure out how we can best position it and, and, you know, maximize its usage. Right. And, and obviously that was a pretty novel use case that open AI came out with a nano banana was a replication of that similar strategy. Um, and, um, Sora was also a replication of that, you know, we'll see if long-term that's successful. So it's, it's a bit part of the core job. Um, secondarily, and I think what you were getting at with the question is that also, um, The, there are no new distribution channels to be reaching people. Um, and all of this first wave of AI has been productivity focused and a bit like bizumer in nature. There's certainly pure consumer fun stuff, like, you know, creating the videos and images for yourself and your friends, but more of it has been on like productivity bizumer type work. And well then, you know, what's the, the best distribution for that, uh, is people talking about the work that they're doing. Right. And whether that's internal companies becoming influencers or then going to content creators and getting them to talk about their workflows that they're using. Right. Um, yeah. Um, like, I don't know if you listened to the door cash podcast, but like all his ads are basically like, I went and use this new Gemini tool to do this big, big, big, you know, complicated thing that, you know, um, the researchers listening will understand. And I mean, the ads are just really good as a result because they walk through the entire workflow and how to use the tool. And so that's, uh, a new emerging kind of style of ad, um, that helps people, uh, you know, reach, um, the right type of customer. And will that be optimal five years from now when we're talking about like, you know, 4 billion active AI users across the world, you know, maybe less so, right. It will feel more like consumer internet again. Um, but right now that's a reflection of where AI is product market fit and has AI has product market fit, you know, with productivity with builders. Um, and it's not, you know, as hardcore, like more pure consumption, um, you know, as, you know, let's say consumer internet over the last 15 years was focused on. Yep. Um, no, totally. I agree with you. And for some reason, this, as much as it seems like a very, very new sort of a growth tactic, it just comes a lot back to the fundamentals. And it's almost, I think the human quotient of what can happen with AI making people talk about what they're doing and, and almost use of, of it as, I don't know, like the word of mouth marketing that was good old back then. And the same thing that's happening across social. So yeah, I agree. Yeah. And one last bit on this, um, the, the whole discovery growth and engagement path that we're talking about is I'd like to understand if pricing henceforth also has a certain effect because of what was happening in the AI space, because I think monetization is a large part of how you retain these people and make them pay for what you provide. And we did speak briefly about how it's so much fluctuate fluctuating right now with, Oh, like that I use as a suite, but then they drop off to use another model or they drop off to use another product. So how's this hitting in terms of pricing of these different platforms? There's a couple of things happening. One is, um, you know, in SAS, um, we thought basically that we had zero marginal cost, so we could give away a lot of value, um, to get into the appropriate user, the appropriate company, and then monetize the team and do all that stuff later. And depending on what type of work you're building with AI, um, your cost can be quite significant to serve free, you know, users. And so unlike let's say, you know, mobile or consumer internet, a lot of the biggest growth stories have not actually been free products, uh, this generation, cause they just would run out of business after like, you know, a couple of months if they were free. And, uh, that doesn't necessarily mean they're profitable or that they're even high margins. Like some of these companies, even though they're paid are still negative margin, and that might be okay. It's that their growth model just doesn't work as, as a free product due to the cost of support. So, um, there are, you know, a few things to think about there, which is like, all right, well, if we look at the last cycle and how we got mobile to many billions of users, we know it was largely free and ad supported. Um, a lot of pundits have basically said, oh, well the age of AI, we're creating so much value that people are willing to pay for it. And I don't fundamentally believe that at all. I believe this generation is going to look exactly like the last generation, largely ad supported on the biggest platforms with premium features and with obviously enterprises having high willingness to pay and all of that kind of stuff. And we're early enough in the cycle that you wouldn't see that because well, a, we don't really have the AI native advertising yet. Um, apparently it's coming soon. Um, I think if anything, the whole personalization is adding to it, anything that feels like it's collecting a lot of our own tastes, even chat, GBD falls for that matter. People speculate that all of this is going to lead, like you said, into the ads. Right. And I mean, look, no one really, even inside those companies knows perfectly how it's going to work yet. They're going to try some stuff. They're going to see what works, but the reality is, is, is, is that the global internet cannot be supported by subscription models alone. And it's a very us centric view that everyone has all this like, you know, excess income to be spending on, you know, internet products. And that's just not as true globally. So if you want to be building a truly global business, you're going to need to be finding ad support. And by the way, that's also true in the U S there's a lot of people that will take the free version and they'll watch ads. Right. And that's the trade off. So I do think that's where it goes long term. Short term, that was not a viable option for startups, of course, because they didn't have enough eyeballs to sell ads and they didn't have the economics to support, you know, free growth. And that's why we've seen a much more, you know, usage space pricing, you know, sort of model for a lot of the companies that are growing really quickly. But I do think that reverts back to the mean of a largely ad supported internet, you know, over the five to 10 year timeframe. And that will, of course, be led by Google and by chat CPT, you know, first, and we'll figure out how some of these AI native ad models are different. And then that will propagate to, you know, other places. But so yeah, so it's kind of like a market entry positioning and then a maturity of the industry. That's giving us a little bit of a head fake on how this is going to shake out. But I, I would bet very strongly that it's going to look exactly like last generation. Yeah. Longer term. Yeah, I totally side with that thought too. And it's also somewhere, I think today's efforts, like you mentioned, when it comes to some of these product launches, or even for that matter, browser launches, right? The fact that there's AI led browsers that's in work, and there's also like hyper personalized means of feature releases or content that's been served is all somewhere leading up to, hey, like, we would show you ads, but then it's going to be like super specific for you. That's what Instagram felt like, or most of even what Spotify felt like about a decade ago, when this model came into existence. And yeah, like the sustenance part is very true, because when they have to think of global markets, that's like one of the options they come back into. And now it adds up as to why someone like a chat GPT or Google Gemini would focus on actually hoarding users and having more usage, because if they chose to go into the ad route, I think that sheer volume is going to help them with the whole distribution part that we talk about, which I think Google already has in some sense. But yeah, chat GPT is like just been building it over time now. So yeah. Yes, quite successfully, but yes. Yeah, quite successfully for at least stopping the consumer chats when nobody was expecting it to be. But yeah, that's that's so true. Wow, like pretty wild times to think of from now on. But yeah, cool. In very similar lines, Casey, I have one more thought, which I'll pick your brain on the LLM search right now everybody's so concerned about, oh, like, has my brand showed up in LLM search and every growth stream is now trying to either invest in a tool or build out their own workflows to figure out how many LLMs have mentioned them or what is it really to work towards it? So what is I mean, if there's a growth person listening to this, and if they are actually working on their budgets, what would you advise them to do and to not do? Like, especially when it comes to the whole acquisition and using LLMs as a means to get to that? Yeah, so a lot of people are saying like, look, the new distribution channel is going to be, you know, AEO, right, like, you know, answer engine optimization. And I think that's largely a trap. For most companies. I don't think LLMs have much interest in sharing their traffic the way Google did, you know, for so long. So I think it's better to view those products as increasingly competitive to your core offering. So you know, for my area, right, which is building like networks and marketplaces. If you know those products historically won by building basically vertical, vertical specific discovery engines for travel, you know, food, transportation, etc. Then like the LLMs have openly stated that they want to build agents that do just that for every vertical, right. So we can debate whether LLMs have a chance of competing successfully there. And I would say generally, if you're good, technically, and the vertical you play in, you know, let's say food is high frequency, I'd bet on you building a better, more specific agent for discovery and transactions than say, chat to PT, you know, or Google. But I think you do need to react to how the largest consumer companies are operating here. And so instead of optimizing to get my brand mentioned AEO, which I think matters for some, you know, higher end discovery stuff. But does it matter for some kind of lower end, you know, higher frequency stuff, I think you want to just be doing what the model companies are doing or planning on doing yourself, which is like, you need to help your customers go agentic themselves. So if you're representing businesses, then you need to be building, you know, the agent version of that business and help them integrate, you know, with LLM tools. If you're building discovery for consumers, you need to be building better, like agentic, you know, discovery, compared to how people are going to be able to do it through chat to PT or through Gemini. And if you're like less frequent, right, where the hope of you owning, you know, that discovery, part of the of the workflow might go away, because like, obviously, the LLMs are gonna take it, then you might need to like move into more of the real world, which is like, well, if let's say I'm not saying this will be the future. But let's say that if you're a travel company, and that chat to PT gets really good at helping you find out what place you should go travel and what hotel to stay at and what restaurants to go to. And you're like, Alright, well, I'll just handle all that for you. I know your credit card information. I know your preferences. I built up the whole schedule, I booked everything. There's a lot of reasons I don't necessarily believe chat to PT is going to be able to do that. But let's say they can, right? What do you do in that position? If you're Expedia? What do you do in that position? If you're the hotel? What do you do in that position? If you're the restaurant? If you're the hotel, the restaurant, you're like, cool, I'll just have my direct integration with chat to PT, they pay me the money. Cool, that's all fine. If you're Expedia, you're like, well, will they book that through? Will chat to PT book that through Expedia? Will they try to go around me and book directly with the hotel the way Google local started to do? And will they even if they book directly? Will they know it's from Expedia? Or will I get like, my brand get like pushed out of that? Yeah. So that's gonna push a lot of these purely like software models into the real world of like, I didn't have branding at the hotel you go on that this was all, you know, done through Expedia and all that kind of stuff. Yeah. So I think you've got plenty of time to figure this out. Because we're not switching to like, all of this being, you know, working really well inside the LLMs by like the end of next year, probably. But over a five year timeframe, this is where these companies are explicitly trying to go. Whether they get there or not, I don't know. But I'd rather bet on them being competent at getting there and then defending against it, like starting like today. So yeah, I don't think feeding into the model companies is generally going to be that strategically advantageous. If it's like, look, I'm buying a high end travel bag, and I want to research the best options. Aeo is gonna be great for that, you should optimize for that. But for some of the bigger like venture scale stuff, they're just not going to be traffic drivers more so than they're going to be people that try to disrupt your business entirely. And you want to defend against that smartly. I think the other general principle that might be helpful to think through is anytime you're thinking about growth, you generally have an asset. It's some defensible data, some defensible inventory, whatever. And if you're going to share that, the question is like, okay, I'm sharing away something differentiated that I've built, which makes it less differentiated. What correspondingly do I get back in terms of growth to make that trade worth it? And so when I was at Grubhub, it's like, okay, well, if we want to integrate with Google more deeply, what are we willing to give them? What do we want to protect? Because it'll just allow them to build our business. And one of the things we said is like, we're going to protect the delivery boundaries of every restaurant. Because if you give them that, then they could theoretically really just rebuild our business quite quickly. And at Pinterest, we had the same question of like, we're more deeply integrated with Google image search. And it's like, okay, well, what are we willing to give away to get that more traffic from Google image search? And what are we not? And we basically said like, well, we can't give them all of the repin data on what the most popular pins are and all that kind of stuff. Because then that gives them everything to rebuild Pinterest. And both of those trades worked out for growth for both Pinterest and for Grubhub. But if you're going to be like, hey, Chad, could you take all my data? And then they're like, cool, I took it, I trained on it, and I'm not sending you any traffic. And I'm not giving you any value, then you're like, well, why did you just do that? You just kind of gave away your business. So I think for certain businesses, a smaller percentage of the people think that trade totally makes sense. Give them all your content, it helps your brand show up, it helps your brand ultimately be purchased. You don't care if it's purchased through chat TPT, or directly on your website or whatever. And then for a lot of us that are trying to build direct relationships with users, and be the real estate where they hang out at or transact, that actually probably isn't going to be a good trade the way it was for Google in 2005. Yeah, no, that's a pretty unconventional advice. But it makes a lot of sense. Because I almost found that it started off as a behavioral shift, almost thinking that, oh, like, let me see if my brand is getting spotted to now taking active efforts to sort of invest in tools or activities that would eventually have the brands listed in these AEO. And to some extent, for me, AEO reminded a lot about going back to the good old days of, I don't know, like answering comment threads on Reddit or doing something that just would possibly do good to your brand and not literally like a tactic because all of that tied back to, in some ways, making the community see a product or making it more discoverable. But I guess now it's just like going a very anti route where you're trying to almost engineer something like engineer or plant something that probably not through the right ways. But then it almost feels like I don't know, you're cheating your game or you're like going around it. And the point you mentioned was very interesting about the competitiveness, like where this could eventually be guarded, but not as open world as what Google's SEO was like decades ago. Yeah. So I think Google made a change philosophically during mobile days, whereas when they launched on consumer internet, they're basically like, it's our job to get you to the best place to solve your problem as fast as possible. And in mobile, because mobile web was really slow early on and because it was harder to browse all the links, I think their users came to them and were like, why aren't you just telling me the answer? Why aren't you just giving me the solution? So they shifted strategically to being like, okay, whenever we're not answering the question directly and having to send you somewhere else to do a vertical search, we failed in the eyes of our customers. And then they started building competitors to a lot of the companies that they used to send us in traffic. They built the travel search product to compete with Expedia. They built a local search product to compete with Yelp. And it doesn't mean those businesses went out of business, but now they find Google as this traffic driver, but also a competitor. And I think LLMs ratchet up that relationship more significantly. So again, if you're a DTC e-commerce brand, nothing changes. I want to rank as high as I can on Google. I want to rank as high as I can on Chattopadhyay. If they offer me options to pay to reach more people, I'll spend money on those the same way I spend a ton of money on meta ads. No problem. But if you were a vertical solution for discovery, you can't say like, well, I'm just going to optimize for AO because it's like, yeah, Chattopadhyay doesn't want you there. They just want to solve that discovery problem directly for every possible problem. And yeah, there might be like a app, you know, call that you can make in the short term, but then, you know, they're going to be like, well, why do I need the app? Now I have the payment information. Now I know all the preferences. I don't need to tool call, you know, Expedia anymore. I don't need the tool called DoorDash anymore. And so me personally, I wouldn't feed into getting them to that future. If I were running one of those companies, there's founders a lot smarter than I am about their businesses, but I don't, it doesn't feel like the trade is worth it. And if you're really a content game, you know, then the content game with Google, which made sense for, you know, maybe 10, 15 years and didn't make as much sense for the last five. It's like, okay, well I give you my content and then you send the people to my page and then I show them ads and that's how I make money. And then Google started answering the questions directly using your content around five years ago. And a lot of businesses like kind of went poof when that, when that happened. And the LLM products are basically just an extension of that. So if you're big enough, yeah, you can get some sort of licensing deal like Reddit or the New York times or whatever. But if you're not, yeah, then they just take your content, they don't give you any traffic. The trade doesn't make sense. So I think you're going to see a lot more people protect their content over the next decade because you know, these companies aren't offering them a reason to share it and maybe they'll figure out a scalable profit share based on content used. But right now they're in not a very strong position to do that and they feel like they don't need to. Yeah. Yeah. That's so true. And they could use all this time to figure out what in that world could their distribution or their tactic look like. And I mean, taking a step back, it also seems to be very different mindset for, let's say a SaaS versus a consumer internet sort of a company, because what I'm hearing a lot in SaaS these days is, oh yeah, like it makes so much sense to make something as an MCP or actually like make it more a tool call within these LLMs or make it more accessible through these interfaces because people are increasingly accepting the fact that someone actually making the conscious effort to go to an interface, log in, do a bunch of actions might probably like reduce in behavior and also to the same point that you mentioned companies like Salesforce and whoever has this, I guess the data heavy platform as such, they're the incumbents investing in their own workflows and their own agents to say, oh, like we, since we have your data, this is like the most best workflows you would end up using instead of a very normalized AI agent. So what this is very different from like the consumer examples that you are bringing through, which is, yeah, like beware of the whole AEO side of things versus a SaaS is also like becoming more conducive to, I guess, the LLMs than what these marketplaces would look like. On the SaaS side, right? So I think there's a few things to think about whether you're incumbent or whether you're a startup, right? And if SaaS is going to be a considered purchase for your customer, then you need to show up where they research those decisions. And that's definitely going to be these tools, right? So you can't opt out of that. And so beyond that, then there's going to be, does, is work being done in these tools related to my business drive better engagement, deeper integration or worse, right? And in the short term with MCP, it basically means people are accessing my data, my tools, my core value in more places. And that feels great. And, you know, a symbol of pretty much every SaaS company's product market fit and something you have to spend significant time on post-product market fit is integrations. Because all of us that use SaaS tools want to use it with our other SaaS tools. And when you get really big, you try to be the suite that just is all of the SaaS tools. But early on as a startup, you can't do that. So, you know, one way to test if a SaaS business has product market fit is if they have a dedicated integrations page. If they do, they probably have product market fit. So, you know, take a company like Notion, right? You know, they have a bunch of integrations. And now that they're really, really large, they're starting to launch dedicated versions of those things they might've integrated with in the past, whether that's calendar or email or things like that. And they can be in a position where they don't have to win those spaces to be successful. They can compete in those spaces and get some extra value, some extra revenue, some better retention. But also if you decide to use another tool, it might be okay as long as you're using it with Notion, right? Yeah. And so that's the difference between like what do I need to own versus what I need to compete for to get, you know, extra value. And for startups, I think the question comes back to like, okay, well, clearly I'm part of a workflow that includes, you know, a bunch of these other tools. And then the question is like, do I have enough of a core that's going to keep me well-positioned in the workflow for my customer? Or does me integrating over the long-term mean that like my part of the workflow gets commoditized into the bigger tool? It could be Notion, it could be ChatGPT, it could be, you know, Microsoft, Google Docs, whatever. And that's the core thing to think about, right? So, and that's going to be a custom question for every one of these SaaS startups coming up that's AI native as to like, okay, I'm feeling pressured to integrate because my customers want it. But does that integration over a five-year timeframe leave me in a stronger defensible position or a weaker defensible position? And if so, I might still need to do it, but I need to do other things to like counteract it. Yeah. Right. Yeah. No, that makes sense, right? Because you invest a lot of time, especially if you're a startup and you invest so much time in sort of a custom integration, but that doesn't scale beyond like what, five to 10 people, then that's not really going to give you like a lot of time. Yeah, not worth the effort for sure. And this also brings me to what popularly I think in the X lingo and LinkedIn lingo is called is the wrapper companies, because everybody who is building something that's easily, I guess like, I guess the taken over or I wouldn't say replaced because now, as you rightly pointed out, a lot of these workflows can be integrated well within a larger incumbent or within another kind of a startup in just a couple of weeks time. It used to be months before, but now it's like just a matter of weeks. So if I'm a founder today and I have to think about the impact of not just the short term, but also the long term, is there something that you would tell folks to think of, especially with respect to, I don't know, like getting copied by or getting sucked into an existing ecosystem of other players as well? Yeah, so I think the key question is, you know, where am I building compounding advantage in my product, in my company, right? And for me, that's a pretty easy one. It's like we're building cross-site network effects. We're helping you discover the best people on certain business problems, whether you want to hire them, whether you want to get on a call with them, whatever. You know, as we get more and more people, even if someone copies the product, they're not going to have the network we have, right? And that's what dominated some of the biggest outcomes in consumer internet, but also in B2B, you know, over the last 15 years. You know, the Shopify's of the world, you know, the Salesforce of the world. They're just, they're actually giant network effect businesses. Those are platforms, you know, I've seen companies like, you know, Pinterest, whatnot, DoorDash are cross-site network effect businesses or, yeah. So if you're a startup, right, and you find somebody that has product market fit, the next question is like, well, how do I sustain it over the long-term, right? And incumbents are way more on top of things than they used to be. They've all read Clayton Christensen. They don't want to be disrupted. It's easier to build than ever before. You know, we all have cloud code or cursor, right? So that makes it easier to copy. So a lot of times your product market fit is just a market entry strategy into opening up some more pieces on the board that can move you into a more defensible position, you know, longer term. So then the question is like, well, what are you moving toward that's going to fortify your walls around your customers, fortify your walls around like revenue and margin and all of that kind of stuff. And historically there's been a few ways you would do that. Which is your classic, like, okay, network effects, switching costs, you know, brand, economies of scale, all these kind of things. And some of these don't feel relevant in an age of AI. Like, wait, how am I going to get to economies of scale faster than Google? Like, that doesn't feel like a realistic conversation we can have, you know, as a startup. Brand doesn't feel like a realistic conversation because brand in software is usually just rooted in the product experience and how delightful it is. And so I would almost like take brand and shift it to like taste. So it's like, well, if I can deliver products at higher taste value, and I can replicate that into more and more spaces, that will probably create some, we can call it rational or irrational preference. Or it's my stack versus an incumbent stack or another startup stack, right? I think Granola is a good example of that, as I think of like, Granola is a good example of that. I guess when Otter and a bunch of these other tools were just ruling over, Granola had that design, the taste that it could have its own. Yeah, yeah. So essentially, if that creates some brand preference there, but that product is a feature in a broader workflow solution we're all using, what are they going to leverage both that wedge and that brand preference into to create a large, defensible, iconic company? I don't know what the answer is, but I hope they make a strong one and execute on it well and have enough time to do it. But that's a great example. I feel like there's going to be many other examples where like, and a lot of this is how Apple has played the game in the last few cycles. It's like, oh, there's a space. Everyone's in it, but their products are kind of crappy. And we just show people what a truly well-developed product experience in that space is, and then we win. I don't know if Apple can do that again in AI, but it's certainly what they did with mobile phones. And that worked out pretty well for them. But then the problem with taste compared to other forms of sensibility is once the taste is shown, other people are like, oh yeah, that is better. Can I reverse engineer it and copy it? And sometimes that's really hard to do. Apple's so deeply vertically integrated with the chips and the design of the hardware and the design of the software, right? That's been pretty difficult to replicate. If you're a purely SaaS business, you can clone the thing. Now you might be cargo culting, cloning the wrong components and not really understanding the core underneath. But a lot of times people that aren't good enough to develop themselves, once they see it, they're like, well, I understand the core elements beneath that made it better and I can copy that effectively. So that's something to watch out for. But I think those would be the sort of founders I would bet on is that, well, if you could do it once, your chances of doing it twice, three times and then creating eventual compounding advantage and defensibility is probably there. Yeah, this is so true. And I think to add on to the copying the taste part, I'd also say that there's one thing to copy and the other thing is to own up what you've copied, right? A lot of people fail at that position where it's kind of easier to say, oh, like, yeah, let's go and build out a linear like webpage, which everybody did at that point in time. But then there's no like owning up to that standard or owning up to that taste. I'll give you the negative interpretation from the last generation, right? So when Dropbox became really large and got to like a billion in revenue, the top three companies were like, all right, that's enough. Apple built iCloud, Microsoft built OneDrive and Google built Google Drive. And did that kill Dropbox's business? No, but they certainly asymptoted after that, right? You know, same with Instagram and Snapchat. They're like, all right, I see your stories thing. If I build that for the rest of the world faster than you get the rest of the world, I cap your ability to get to a billion users, right? And they did do that effectively. So the incumbents are smarter about doing that well and they're just doing it a lot faster this cycle than they did last cycle, right? That's true, yeah. So, you know, Glean hits a hundred million in revenue and everyone's like, okay, I need to stop that now. I'm not going to wait until you get to a billion, right? Yeah. That so far has not stopped Glean revenue growth, thankfully. But yeah, the companies are just like, all right, I see the wedge happening a lot quicker now. I can react a lot quicker now. I'm incentivized to react a lot quicker now as well. Right. And the resources to get to that is also like a lot quicker now because of access to a lot of these tools. So yeah, very, very contextual in that sense. And to like, come and zoom into Superme, for example, I love how we're talking about the whole AI slot, but there's also this complete human knowledge and how you're making LLM access this human knowledge with Superme. I'd love to understand more around like what prompted you to build this and what exactly has been the most, I guess though, what turned out to be surprising for you as an artifact, which gave you like super high signal because you're building these profiles with so many tons of high operators. So what's been the most surprising turnover for you from all of building this out with all of them in the circle? Yeah. So I think some of the things that we saw that, you know, made us want to jump to this is that a lot of sharing of like business knowledge between peers has moved off of the public internet over the last 10 years to like private chat groups or slacks. So it feels like the best knowledge on solving business problems has moved, you know, increasingly away from something that, you know, a normal person can find. And, you know, the reasons for that, right? Like X or LinkedIn, you know, they rewarded more, you know, viral content more recently, less nuance. Um, if you share something, you know, that's a little bit, um, different, you might get a bunch of like harassment, right? So that's like moved it. Um, and then when you couple that with like, well, now these LLM products, uh, have are indexing all the publicly available knowledge, but new business insights are not being shared with the public internet anymore. You're like, well, where does this mean for our ability to kind of, you know, uh, make the best business knowledge, you know, transparent, right? And, um, the other thing that, that bothered me about the shift is that, um, as we switch away from Google search into, you know, LLMs, it moves from 10 blue links to an answer and answers are like really great for facts, but in the business world, we prefer to hear multiple perspectives and then make up our own mind. And so you're kind of losing multiple perspectives as blue links, you know, shift to answers. Um, so what we were trying to do is say like, well, how can we date back to indexing the best cutting edge knowledge from the best business operators and how do we make it easier to view the multiple perspectives so that you as a founder, you as a product manager, you as a designer can make up your own mind with all the best information. And so that's how we designed, you know, super me, you know, upfront, I think what surprised us about the usage is I think we anticipated, you know, the first version of super me is you would come, you know, would ask your profile, you would ask my profile or you would ask search and you would ask, you know, sort of a generic question like, Hey, how should I compute my activation rate? Or, you know, Hey, uh, how do I understand if I like have liquidity in my marketplace? Right. And you'd get an answer that's going to be sourced from my content. So you're going to know it's not hallucinating. If it's something I don't know anything about, the AI provost can be like, I don't know anything about that. You know, it's not going to hallucinate. Yep. Instead, what happened is people would come in and they drop like three paragraphs about their business. Um, they'd be like, I'm working at the series beef in tech company and I have this problem and here are my metrics. And when that happens, what our product can do is, is go agentic and delivering the solution. It cannot just say, well, here's how to calculate an activation rate. They can say your activation rate metric should be X. Here is how to calculate it. Here are the things you probably need to do. Given the context, I understand about your business to improve it specifically. And we just didn't think people were going to do that early on. We thought that was like something that would merge two, three years down the line. I think we failed to anticipate is that all of our users are early adopters of LLM products. That's why they're trying us. Yep. They know better prompting means better answers. And so we're giving a lot much more personalized advice to the AI profile than we anticipated. I think what surprised us on the, we'd call it, you know, the supply side, the people that are getting a bunch of questions is that we designed it in a way where like the AI profile is all always going to answer on your behalf and, um, you want to check in on it and make sure the answer's right. You know, chime in every once in a while to improve it, to train the model the next time that question asked. But what's happening a lot of the times is there are people that used to get too many emails, too many questions on social media, et cetera. So now they're sending people to their super me's. And what we're finding is that they used to be like, look, I would just would never have gotten to these emails. But now they're like the fact that the AI answers on my behalf actually means I can chime in as well because the AI has done 90% of the work. Right. So it allows me to put a personal touch on way more, um, of these, uh, you know, way more of this outreach than I could in the past. And that I didn't really anticipate and how people were going to use the product. Um, that's been really cool. Um, also to see. Very nice. I know you folks had a beta launch very recently. Like, is it open for other folks to try globally? Yeah. So it all works self-serve. You can sign up, you can create a profile, you can, you know, search, you can have conversations, all of it's available. Right. Awesome. I'm going to like link it up in the show notes. Would love for folks to check it out for sure. Um, Casey, thank you so much. This has been exciting and I had such a fun time discussing and in fact, dissecting all of what's happening around in the AI space now with respect to growth. And I think otherwise as well in terms of how these business plans play out in the longer run. So thank you so much for being here and sharing your thoughts with us. Yeah. Thanks for having me. Awesome. See you.