They can't afford chat GPT to say something that could honestly be, you know, very incorrect, but they will also make it sound very factual. Like, they're very good at making something that isn't right, sound right, and you can't have that at an enterprise. So what we realized is, okay, well, we need to constrain this, right? We need to constrain the generation part to something that they have. And at this point, all the research pointed towards the amount of content that these brands are creating on an annual basis is just increasing at an exponential rate. Like, content is being pushed out quicker than ever before. And part of that is because of the content generation, you know, applications like a Jasper, which leverages GPT. But ultimately, what we're seeing is, okay, so we have a ton of this content that's already been approved by the brand. So let's configure this large language model to be able to actually look at this content, ingest it, and then use that to answer the questions that a user may have when they go to that digital property of the brand. So what we built is, you know, the initial layer is this platform that allows the AI to go in, ingest content, you know, whether it's a knowledge-based content management system, documents, wherever it is. And a lot of these, you know, large enterprises have it. I mean, their content is everywhere. And so what you want to do is take in all that content, and really allow the AI to be able to parse out what is the best answer based on the user's query. And that's leveraging these large language models. And then, you know, that's kind of where we, all right, well, let's push the envelope a little further. What if we were, you know, how containing or maintaining contextual awareness is very difficult in its manual? Can we figure out a way to allow the AI to do this from an automated fashion? And so then what we did was we built a layer on top. And this is patented is the ability to, once the AI ingest that content, what we're doing is we're clustering all these similar topic areas. And once those topic areas are clustered as far as the similarity between them, like semantically, then when a user can go in and ask questions, they can ask those follow-up questions. You know, they can say, hey, you know, tell me what is a traditional IRA, you know, tell me a little bit about that. Okay, how does that differ from a Roth? Okay, what are the contribution requirements? And being able to go through that conversation while maintaining context of the fact I'm talking about that traditional IRA. And so, being able to do that automatically, you know, is a huge step as far as being able to have that conversational experience. So now we've connected to their content. We've now made it very conversational and the fact that I can just continue to ask these questions.