How AI-First Companies Differ from Traditional SaaS (Full Transcript)

[00:00:00] Speaker 1: What do you think are the differences between running an AI company, AI first, like a native company, and then traditional SaaS company? I think it presents a lot of like interesting strategic questions that I feel like we're running an AI company, right, as opposed to running a SaaS company. Like, where in the stack do you play, right? Are you building true foundational models? Do you take true foundational models that's open source and make them your own through lots of intensive, heavy R&D? Do you just use APIs or like wherever in that stack you kind of sit? I think you have to be intentional about that, but also accept that it can change over time, right? So when we started the company, there were nothing, right? There was like literally nothing to go from. We had to build everything ourselves from like the facial landmark detections, for example, which was like a big important part of like the first iteration of the technology. Now in a world where like there's so much stuff that's happening online, we're shifting to all these like generalizable models. A lot more people are just thinking about AI video, experimenting with AI video. And so I think a lot of like the hard questions that you face is this thing of like, when do you chase the shiny new thing, right? Where like, okay, clearly this is going to be the future of the space. Would you just like jump ship to this now? When do you kind of hold your guns? And I think we've definitely made mistakes in both ways, but sometimes you've been too eager and sometimes you've been too conservative. And because there's so much uncertainty with these things, right? I think that's like a bit different in SaaS where I think if you're building a FinTech company, like if you can imagine it, you can probably build it. And that's not to talk that down, but for something like this, right? It's like, I want to make videos of two avatars sitting in a chair like we are right now and be able to fist bump, right? Like, yes, that's going to be possible someday, but there's a lot of fog of war in terms of like, how exactly do we get there? Whereas if it's like a FinTech thing where like we want to support crypto, there's another set of objectives, but you can lay them out a lot easier, which means you can plan for it better and probably also means you can execute on it faster. I think also like what is the primary value driver and why is your company valuable? I think, you know, when we started out like the first couple of years, probably most of the value for us was like the models for the build, right? The technology in itself. And I think where we are now as a company is that the AI models are definitely a big and very important part of like our value as a company. But we've built so much stuff outside of it, right? All the boring stuff like permissioning, being ISO 42,100 compliant, like the publishing platform, like all these other things together. They are very valuable because we've built the workflows, right? I think for a lot of companies, the value will be in the models and that's like, that's the right thing to do. And then you need to make sure you're really, really winning on the models because that's what you're betting your company on. I think for some like us, I think probably the value of the models should decrease over time, right? And that's not because we don't want to do like very deep research, but we build the company in a way where the value should accrue from the totality of the platform, not just from the models.