Startup founders confess their toughest challenges & nightmare situations...

I'm going to be talking about AI, but it's probably not what you think. I have not seen anyone do an episode like this. I'm going to bring on AI founders. What does that mean? Founders who are working with AI and we're going to talk about their challenges and then we'll also have questions that folks have submitted about AI and their thoughts about it and my thoughts about it. And you get to hear all that regardless of whether you're working with AI or not today, I'm sure AI is in your head. So I think you're going to have a great time with this episode.

All right, Rapid Fire round. This is like not my favorite round because it. I have to answer for myself and personally, but we're going to do it. So rapid fire. I'm going to answer these questions as fast as I can because I really don't want to answer them. Have you been tempted to start an AI company? If yes, say more. If not, why not? No, no, no, no. That's not the way I think about it. I have companies. Am I going to be using AI for those companies? With those companies, are we going to use AI in our marketing because we have AI or whatever? Yes, I. Absolutely. Why? It's the tailwind. It's what every. It's on everyone's minds. And ironically, I was just talking to my co founder and a bunch of our team about this very topic and how we are going to use AI. So yes, but it's not tempted to start an AI company. I don't even know what an AI company is, to be honest. These are software businesses or hardware, if that's what your choice is. What's the biggest false belief people have about AI? In my opinion, it's the fact that AI can solve every problem. So the way I think about it is people think of AI as a hammer, let's say a sledgehammer. And when you have a hammer that powerful, everything's a nail. But that's not true. You really need to dive into the real weeds of solving problems people have with AI instead of thinking that AI can solve every problem and it's always about customer value and problems first. So I think the biggest false belief is AI is some magical thing that will just instantly make your product better, make your business better, or what have you. I want to start an AI company. Have I missed the boat? You want to start a company. You have not missed the boat. It's never a bad time to start a company. Sometimes some people say it's better than Ever and this and that. It's always a good time. The one thing I would say, though, is don't start a company, any kind a startup, a business, whatever, unless you really understand what it's going to take. And what it's going to take is not this whole, like, it's always fun or it's always exciting, what it's going to take. It's going to make you question so many of your assumptions about not just business, but life. And so start a company because you want to learn to love how to solve problems and do it at such a pace that, like, your mind probably explodes. That's why you should start a company, because you want to learn as much as possible as fast as possible. It's probably the number one way to learn about not just business, but life. How far away do you think AGI is and will it just make every startup redundant? I don't know. Frankly, I don't care. We should go ask Sam Altman and get him to tell us the truth. Maybe kidnap him and ask him. I don't know. So, really, like, this question is existential and that's why I don't have an answer. And it truly should be existential. And if you talk to, like certain folks that are AI experts at Facebook, they've recently said, oh, it's really far away. If you talk to Sam, I believe he said all kinds of different things about it depending on kind of what his opinion is at the moment. And I don't even know if he knows. Can we out compete intercom by leveraging AI? We're building customer communication platforms for SMEs. Our competition is Intercom. How can we compete with giants in this market by leveraging AI? So the first word that comes to mind with this question is differentiation. If you can't be different than them, it's going to be much harder to compete. That doesn't mean you can't. So first be different, which essentially means find a way to be better. Better can be better for everyone, or it can be better for a certain segment that Intercom is sort of not paying attention to. And the way you would figure that out if you went down that path is basically talk to people who have either used Intercom or have churned from Intercom and try to learn more about why. Once you do that, you're going to get a much better idea of how you can compete with Intercom. If you're going to decide to compete by slicing up their customer base and finding customers who are, for you, opportunistically not very happy with Intercom and match the differentiation that you can provide them. So that's kind of the tactical one way to think about it. If you're willing to do things differently than Intercom when it comes to going after a different customer and even also figuring out who that different customer is and what you can do for them. The second part to this answer that I would say is, let's say you don't want to do that and you want to hit Intercom head on. There's only one way to really do that well, which is you have a distribution advantage or you have a cost advantage. So distribution advantage means you can get customers cheaper, which is related to a cost advantage, which could be that you can service customers cheaper so you can charge them less than what Intercom charges them and still make a bunch of money. So if you're going to go direct, it's really about distribution and having a better path for distribution. If you don't have that, I would highly recommend that you find gaps in their product, gaps in their customer base where there are customers that you can service better than them. With differentiation, how can you tell if your AI product is actually doing well? We got 3,000 users in our first month. This would be a good sign for a SaaS product. But our churn is high and not many people have converted to paying users. What signal should we look out for that our product is working? Huh? So the reason I say huh is because the question is what signal should I look for that our product is working? Well, I think you have signal that your product's not working, so you're not really looking for signal that it's working. What you're looking for, and if I would read in between the lines, is that you probably have some customers that are actively using your product. You might have some customers that have converted to paying users. Like you've said, even if it's a few, you want to figure out what makes these customers different. And this has nothing to do with the AI product. Nothing. 3000 users in the first month for even SaaS products. If people aren't buying, and it's a freemium tool, that's normal like, and that doesn't mean anything. There's Nothing special about AI. Sure, the AI hype might have got you those 3,000 users, but a lot of people are willing to kick the tires with software in general. AI just happens to be something where you can get that kind of user base pretty fast, pretty easily. As sort of, I think you're pointing out so my bigger kind of piece of advice here is step back and figure out what the value is that people are getting from your product for the people who are getting the value and see if you can find any signal there. Few things you can do here that are very tactical. One I'll mention right when people sign up, send them an email. That's very quick and simple. Make the subject say quick question. I know it's played out. I came up with this a long time ago and a lot of people have used it and just simply say, I am the founder, we built this product and all I would love from you to do is give us some feedback. Please answer this one question. Why did you sign up for the product and what were you expecting from it? And I would do that even if people have churned or whatever. As long as you have their email address and you can email them if you have 3,000 users, I don't see why you wouldn't get at least 30 responses to that question. That'll give you more directional idea of what they expected when they signed up. So a lot of AI products, they just don't meet people's expectations. That's why they have a high churn. I suspect that's what you're dealing with. Should we do a freemium model with our AI product? We're building a tool that generates title ideas for YouTube videos. The problem we're finding for each user cost us around three to four dollars per month. Is it worth offering a freemium model or how would you think about pricing? When I hear that your product costs three to four dollars per month, I'm instantly like, you cannot provide the product for free. It just doesn't make sense. What you can do though is you can provide examples of existing YouTube videos that are out there and how your product would generate titles for it and use that in your marketing and basically either proving that your product can do it better or or proving that your product gets good enough headlines or as good headlines as what's already out there. Like imagine taking Mr. B's top five popular videos and basically using that as part of your marketing. So I think freemium doesn't necessarily have to be just a tool that people can use that costs you money. Free tools can be things like what I just described. And when you think of it that way, those might just be a one time cost for you, not necessarily an all the time cost, like a running cost to provide the product for free. So on another note, that's related to this, if I really thought that My product should have a free plan because I think I can get more users or I have the distribution. So that's a key thing. If you have a free product and it costs you three to $4 per month, but you're able to make more than three to $4 per month because the free usage creates some kind of revenue or something for you or value, like more customers coming in or whatever. There is something to be said about that being a marketing cost and chalking it up to that. So I'm not saying you should do that, but that's another way to think about this problem. So the real advice I have is just think deeper about it and think more creatively. All right, I'm going to be talking to some founders about their biggest challenges and opportunities while they're building companies using AI, we can jump right in. Tell me about the product and what your challenges.

So I run a startup called Coso AI and what we do is we use artificial intelligence to automatically create social media content for brands. So we've been building it over the last couple of years and now we're just kind of launching the product kind of. At the moment We've got about 15 paid pilots active with different companies and just getting user feedback. I guess our challenge and how we're trying to. What we're trying to figure out is like how to think about the different ways to integrate like LLMs and generative AI into the product. So like on one hand you've got, obviously you can just train your own stuff from scratch and then on the other side you've got the other extreme where you basically just use out of the box open source stuff which can get you to market quickly. But then you don't have much of a technical differentiator compared to other companies. But then is that even, you know, in a year's time the open source stuff might just be better than anything you can train yourself now anyway. So that's just waste of dev time for a few months if you do that now.

I think it's super easy to forget with new technologies like this. And I saw it happen in crypto, I'm seeing it happen in AI where like we get so excited about what's possible and like how it can be possible and get kind of caught in the sauce, like lost in the sauce as they say. So where I'd be actually focused is how can you create content that is so spot on with your AI that someone is just excited to post it? And that depends on a lot on who your customer is. Certain things you have to do, whether it's like include their brand voice and all that kind of stuff. I don't think I would recommend from a customer standpoint trying to go spend any effort on tuning anything or training or anything like that. What I'd actually be doing is, is aggressively doing what the engineers call evals. The reason I'm saying that is contextual to your product. You need to provide people with content that they love, that they'll just hit post without any edits. That bar is already pretty high. Like I can't just do that out of the box. I have never seen it do that. We have to think of this almost like a maturity phase. Like what phase are you at in the maturity of implementing AI into your product today? It sounds like with where you're at, you know that you want the whole business to be powered by AI in terms of the product and the experience. You have those 15 customers or pilots or you know, early access, beta, whatever you want to call it, customers. So I'd be just trying to make sure that for every single one of them you understand whether you are able to produce great content with the current methods you're using that are low hanging fruit. And to me it's okay to. And this gets into the kind of the investor side. It's okay to tell an investor, hey, we are really good at testing and evaluating and prompting to the point where we're very focused on this one thing for this stage. Can we get AI regardless of how the AI is done, to produce content that someone is willing to post right away? Because that's already the biggest challenge to figure out. And if in that process you learn that you need to train or fine tune or do rag or whatever, you will know the answer. But you're at a pre seed early stage, you've got enough traction it sounds like, to raise a little more money to buy more time, that there's no reason to go into the hardcore kind of deeper, deeper down the stack sort of training and things like that. So with James, the problems he was describing are sort of becoming the classic conundrum with AI. Now where do you use the APIs or do you go deeper and do a bunch of fine tuning rag, even training and things like that. The fun thing is all those things are really easy to do now you could go to aws, GCP or Azure as these cloud services and a reasonable engineer can actually figure out how to do that within a week, let's say or less. There's also a lot of APIs and tools to do that, whether it's together or, or open router for some of this stuff to test models. But at the end of the day what matters and what I wanted to get James to get to was what's the customer value? And how do you define all this around the customer value and how do you build systems and feedback loops which I think oftentimes with AI like people just slap the API on and call it. But really if you focus on customer value, everything else can be solved whether it's with AI or not. As long as you're understanding what is the value you need to deliver, how can you essentially test whether you're delivering that value? And then how can I help you? You notice I used AI at the end there because AI and customer value don't always go together. Just because we got this shiny object with AI and we can build better software and better experiences for people doesn't mean that it's going to solve every problem. At the end of the day you have to go back to customer value. Let's jump right in.

So I'm the co founder of Dial Me and Dial Me essentially helps businesses build voice enable agents. We launched this back in September last year. So it's been like four months and we're trying to figure out where the demand was. And so far it seems like we've gotten market pull in different directions. Like people in HR want to use it for pre screening, people in marketing want to use it for market research, people in sales want to use it for inbound sales lead qualification. And we're in a bit of a dilemma right now because do we stay broad like a type form or Qualtex or do we actually build for a specific vertical because each group wants a different set of features and actually building the super set of all those features would add complexity for everybody.

Thanks for kind of shaping the dilemma. Good news is this is a dilemma. Outside of AI this is a dilemma. As you mentioned Typeform and others Qualtrics everyone has had at some point and you'll notice there are Qualtrics and Typeform competitors that are much more niche. Right on very specific use cases. Whether it's things like only employee surveys, right all the way to only health related vertical. And so the one thing that comes to mind for me is both of those things are different types of product marketing. You have to do. Are you going to go to a HR event if you are a horizontal product? The answer is no, not likely. Are you going to put out templates that work for HR if you go horizontally, probably. So a lot of that, what you're good at, what energizes you, you should know what you're getting into with those two different options. So I would spend some time understanding the difference in the go to market, the marketing, the sales process and things like that, and what you folks get energized by, have experience with or want to learn. That's thing number one. Thing number two, which might be more important than thing number one, frankly, but maybe not, is where's the money? Because if it were me, I'd be like, no, no, wait, hold on, where's the money? And it's not likely in the horizontal thing. It's most likely in some vertical. But then you would be committed to finding the right vertical and going deep into it to your point about the features you're hearing. So then I would, I would actually literally one. One factor and one question I would ask all of your customers is if we did those things you want, how much would you pay us? How much more would you pay us? Who else would be involved in the company for those things? So I would be doing a lot more customer development to understand the viability of each of the verticals and whether you should go deep into them or not. And if you kind of hear a consistent like no one's willing to pay more money for those features or whatever, you might want to decide to go horizontal and not vertical because you can then build a more generic tool set and get away with it because the price sensitivity is out of whack. So Giri's company is super interesting because we didn't get to talk much about AI, but a lot of companies that are working with AI have this problem of do we go horizontal, do we go vertical? Because the tooling can be used in many different scenarios. And really what people need to think through is what are you committed to? Are you committed to a certain type of company like a vertical because you think they're going to make you the most money? Or you have background in that? Or are you committed building a much broader product that turns into sort of a suite of tools that can work for anyone? Because in both cases you kind of need to build a lot of stuff. Thanks for listening today to the Heaton Show. To apply for the next show to come on it, go to Heaton Show.

The show is produced by Supermix. Thanks for joining.