A Product Market Fit Show | Startup Podcast for Founders
00:20:31 · Alex Halliday
[00:20:31] S0: Yeah, it's a great question. I mean, you certainly could do that. It actually might work quite well, to be honest, given that use case. But if you think about content in the current market, competition exists on the startup side but it also really exists on the content side as well. So each piece of content is in composition with a sea of other pieces of content and so to get the most from an AI model. It's much less so the capability of the model, it's much more so the context that you can have available for the model, right? When doing any sort of content task inside of AirOps, we have a huge data engineering team whose sole job is to make sure that we can give that model an unfair advantage when it's tackling your problem. So if it's, for example, a page you want to have ranking in Google search or it's a particular question for ChatGPT. Before you even start writing we understand every single page that is in competition with it, we understand all of your pages that could be in competition with that, you don't want to cannibalize your own content. We understand which questions you should even be going after in the first place. We collect almost one point two billion answers every sixty days from these answer engines and we analyze them. And then once you create your content, we also have a full picture of your tone and voice, your do's and don'ts in your language, your sitemap for internal linking. So you can think of this as like if Claude is the model at the beginning in the center of the picture, can you give it these concentric circles of knowledge that are actually quite hard to compile, but give it that unfair advantage so the content you create is more likely to win, and then the nice thing also about doing things through us is we also track the impact. So you know, hey, in January we published twenty eight pieces of content. We know that ten of them are absolutely crushing it and we should go do more of those. And that knowledge feeds back in to our agent on a go forward basis. So I think in the case of like working with your transcripts, you would have a really good result just doing that with Claude Code and that works well. I think when you're dealing with tens of hundreds of thousands of pages or millions of pages.
00:22:33 · Alex Halliday, Pablo Srugo
Prioritization is the most important thing, followed by the context you can provide on competition and those are two things we nail for our customers to make sure they continue winning. [00:22:44] S1: I'm really worried because listen, you've been listening for what? Ten, twenty, thirty minutes now? Clearly you like it and the thing is, the next episode is way better and you're going to miss it. You're going to miss it, because you're not following the show. So take your phone out and hit that follow button. And do you see any reasons for what works broadly? Because again, we think about this strategy, it's the same strategy just done with AI. But is there any reason that this changes? This is something you hear a lot. It's like, well, there's going to be so much AI written stuff, slop. Whatever you want to call it, whether it's good or not. It almost, when you just think about that at a high level, you're like, there's no way this is still the future. There's no way five years from now, it's still all about blogging. There's going to be so many blogs, but is that real? Or is this just going to be, this is still the strategy and there's a reason for it?
00:23:27 · Alex Halliday
[00:23:27] S0: I think the human web will continue to exist. I think AI content gets conflated into one blob. I think some of the best content created uses AI and much of the worst content uses AI. That's not the right distinction. So I think understanding that there's really great applications of AI to improve content quality and also improve content health in the long term, is something that I think folks need to understand on that front. But the human web is designed for humans to click around and right now, we're duct taping agents over the top of that. So ChatGPT is going to a search index and is retrieving pages, and including them in its response. And there's a lot of protocols out there that look to try and unseat that, including one called NLWeb and, obviously, traditional MCP as ways in which ChatGPT can ask us questions about the Product Market Fit show and get back the perfect snippets without ever having to visit your website. Irrespective of what the protocol is, whether it's going through a search index and going to the web or going directly to an endpoint for the Product Market Fit show in this example. The underlying content needs to be really, really good. So I actually look forward to us moving on from creating yet more pages of content to something smarter and we actually interviewed Glenn Coates from OpenAI at our last conference. Who is running the OpenAI apps program and he said they're really actively exploring this for later this year. And the way he positioned it to me was you want to be the sharpest tool for the job. Meaning if ChatGPT wants to know about congestion in Barcelona or the best restaurant to go to in Copenhagen. They are going to start to form an opinion of which tool is the best for the job to answer that. Which can give the model the best quality response and allow the model to do even more for that user's journey. Yeah, it's going to change. I think as we've seen with many of these things, the behavior of going to websites and reading content is such a muscle memory for billions of people. The web isn't going away in its current form, but it will have a sibling of kind of agent first web, and I'm excited to see that develop.
00:25:33 · Pablo Srugo, Alex Halliday
[00:25:33] S1: When in the story did you land on the current idea and maybe share the story of how you went from what you were doing originally for marketers to what you're doing today? [00:25:42] S0: Yeah, so I think my co-founder and I got us to about $1.3 million in ARR selling ourselves. [00:25:50] S1: Just the two of you? [00:25:51] S0: Just the two of us. Yeah, and it was a slog. You're on the road, you're trying to persuade people that you can meet their problems. You don't understand the problem well enough to be honest at that stage. It was very hectic and you're also doing all the deal paperwork. It is full on product roadmap, design, everything. [00:26:08] S1: So before you move on, why did you keep it to just the two of you? I've heard a few people do that, and I actually think it's a good idea. But for you specifically, why did you keep it to just the two of you to get to a million ARR? Why not hire at least, I don't know, four or five more people? [00:26:19] S0: Sorry, I meant two of us actively selling and dealing with customers. Yeah, so we were doing the full deal lifecycle, prospecting, we were doing everything. I think the team was probably about eight or nine at that point, yeah. [00:26:32] S1: That makes more sense, gotcha.
00:26:32 · Pablo Srugo, Alex Halliday
[00:26:32] S1: That makes more sense, gotcha. [00:26:34] S0: Yeah, yeah, yeah. It was still a knife fight and I remember right around the time we raised our Series A, we had about sixty percent of our revenue came from these core use cases. It was larger B2B SaaS businesses, their content was stale. It was really bad, they had these blogs they've invested years of work in and all the product positioning was out of date, all the links were broken. They were like, oh, this is horrible. Let's contact AirOps and we'll help you with that. The real inflection point for us was actually hiring our first sellers and actually having them, have to learn and then practice and then perform the whole sales process without us. And that handoff was so nerve wracking. But what it forced us to do is really kind of focus our solution and what we were talking about. I think that transition, you shouldn't do it too early. You definitely shouldn't do it too late. But you also need to make quite a lot of decisions and stop doing a lot of things in order to make a fresh sellout coming into that environment successful. So it's a forcing function for focus. [00:27:37] S1: Give me a sense of just that growth curve. How fast did you get to a million? How fast did you get to, five or ten or whatever?
00:27:37 · Pablo Srugo, Alex Halliday
[00:27:37] S1: Give me a sense of just that growth curve. How fast did you get to a million? How fast did you get to, five or ten or whatever? [00:27:42] S0: We got to 1 million in Q1 of 2024, and then we hit ten in 2025. That was kind of a big ramp for us. We actually hit thirteen in Q4 of 2025. So that just was like an insane year for us. It was the machine starting to work, right? It was the positioning getting better, as I mentioned, and just it really clicking with the market with the tailwind of AI and people being concerned about ChatGPT, and shifts there. But then also internally, really learning how to have a repeatable motion. Which something that my co-founder actually has really been leading, but it has been a masterclass to me in iterating on process and just continuously practicing, and tweaking as we learn more and more. And the sort of sophistication of that just building over time. [00:28:29] S1: Yeah, let’s go deep on that. I’m thinking in my head, I’ve got this portfolio company I work with closely. We’re at $2 million ARR, things are clicking, product market fit is there. We’re trying to get to that 10 number, and things are just working but it’s a little bit still happenstance. Still a little too messy, not predictable. What did you do that really worked? What is that setup that you ultimately ended up with that helped you grow from one to thirteen and whatever that is, eighteen months or so?