00:04:54 · Dan Lee
[00:04:54] S0: Yeah. There's there's a lot. So for example, we've started with a lot of the most advanced companies in the world using Nooks. So, yeah, over 2,000 companies from AI native startups like Cursor and Notion to enterprises like Toast and ServiceTitan are using Nooks to augment their team... Their sales teams with agents. But then there's... You know, a large part of the market is, you know, maybe less advanced or less AI pilled companies that can still get probably even more value from Nooks. So, you know, a large part of the investment is, like, accelerating growth in these verticals. We are selling into new geographies. So, you know, we started, you know, mostly in in The US, but we're opening offices. We're opening a New York go to market office in UK and Ireland and Australia and other parts of the the world. And a really big focus is on new products. So where we started is we started by building tools for sales reps. Think like calling, emailing, doing research. We translated those tools to also be used by agents. Mhmm. And we're expanding across the entire revenue life cycle, where we started a lot with the top of funnel, where we can show really quick results. And now, though, you know, doing work across inbound where we have agents that can, you know, hold full conversations in with the products. We have deal agents that, you know, can basically manage all your deals for you, understand, like, what's slipping through the cracks, and, you know, making sure that you're on top of all your follow ups, and even, like, renewing and expanding accounts. So a a big investment is going across, like, expanding the product suite. And then the the last area, I'll say, is in applied research. There are some really hard, like, kind of human agent collaboration problems that we need to solve. Because sales is more about persuasion, you know, we're doing things like fine tuning our own open weight models on, you know, Nooks proprietary data that outperform Claude and ChatGPT on sales specific benchmarks on things like cost and quality and latency. We're building actually, like, RL environments for sales because, you know, because sales is a lot less verifiable. We're building, like, basically, these simulations where you can simulate what a customer... Like, how a customer might react.
00:07:08 · Dan Lee, Ruby Justice Thelot, S3
So you can actually test, like, hundreds of scenarios to see which one's gonna perform optimally. [00:07:14] S1: Man, this is really interesting to me because unlike something like coding or even law, which law is tricky, those are a lot more verifiable than something like sales, which is a very human persuasion. You have to understand that person. Do you think there is a future where AI can do that better than a human can? [00:07:34] S0: So our agents have been trained on, like, billions of interactions. And I think... So I'll I'll give some examples of things that we've done to optimize this problem, and then I'll talk more about the feature. So, like, in coding, for example Mhmm. When a coding agent is generating code, it'll pull in relevant parts of the code base, and it has some function to determine what's relevant. In sales, when you're trying to prepare for a meeting or, you know, send an email to someone, you actually wanna pull in relevant sales scenarios. But then there's this question, like, what's a relevant sales scenario? Have you... You know, or is it going after the same persona? Is it... You know, they've had similar objections? And we... That's an interesting, like, kind of hard technical problem that we've had to solve in building our own agent harness for this. And then there's also... Nooks sits at this unique position where we own the layer between the sales rep and their end customer. [00:08:33] S3: Mhmm.
00:08:33 · S3, Dan Lee
[00:08:33] S3: Mhmm. [00:08:34] S0: So we're able to see all the decisions that a sales rep makes in their in their process. We're also able to see how... All the ways that the customer reacts. And as a result, we're able to... You know, we have this active learning process to learn the the rep's decision making policy, so, like, mapping from scenarios to, like, discrete action sets. And we also are able to build these agents that simulate the customer. So, you know, that's kind of where we're today. I think over the longer term, like, there's just been less data on persuasion in, like, the training sets for frontier models because persuasion is all about understanding what people think, and, like, that's not observable. So, you know, the the best you can get is this kind of sparse and, like, delayed feedback loops. And sales, I think, is the richest dataset on persuasion because it's the most concrete business application. But yeah, I think that models are gonna become a lot better, not just at reasoning, but at persuasion, which is gonna change sales.
00:09:29 · Ruby Justice Thelot
Ad[00:09:29] S1: I wanna shout out our sponsor, Lovable. You know, the app you've been meaning to build, The internal tool, side project, or product you'd otherwise lose a weekend to? With Lovable, you get software you can ship today. That includes an editable code base you can inspect and change, plus two way GitHub sync. Lovable also handles the backend infrastructure, including managed Postgres, auth, storage, hosting, payments, and more, so you don't have to wire it all up yourself. And through Lovable's MCP server, you can create and deploy your project directly from the agents you already use. Turn ideas into software people love with Lovable. Lovable. Dev. Now, onto the episode. What does that mean now? If models are a lot better at persuasion, just zoom out. How does that impact the landscape as a whole?
00:10:15 · Dan Lee, Ruby Justice Thelot
[00:10:15] S0: It's really interesting. Everything is sales. Yeah. Right? Sales is about understanding the person across from you and, you know, understanding what they're thinking. You know, sales is recruiting. Sales is raising capital. Sales is high... You know, like, honestly, I'm I'm selling all the time and, like, know, hiring candidates. So I think there are a lot of application... Actually, so one... We were talking earlier about, you know, our our backgrounds at Stanford. In in 2020, I was actually gonna do research project at Oxford, a study abroad program, on emotion detection Yeah. Because cameras are getting really good. You can see, like, you know, your pupils dilate or, you know, your heart rate changing your neck or, like, capillaries in your cheek flush. And can you actually understand, like, you know, what is someone thinking? Right? So if you ask, you know, Donald Trump, is he gonna, you know, go invade x country? He probably wouldn't tell you, but, like, he might react. Yeah. So, yeah, like, there's all these interesting applications of, like, you know, kind of this human AI, both, like, collaboration and, like, understanding that I think the world is gonna look really different. [00:11:26] S1: So you brought up an important point. Everything is sales. I think often when you think about sales, you think, okay, a product, and they're trying to make money. But everything is sales, hiring, recruiting, conversations, a little bit as we're talking to the audience, whatever it [00:11:40] S0: might be.
00:11:40 · Dan Lee, Ruby Justice Thelot
[00:11:40] S0: might be. [00:11:41] S1: What does a world look like when you can use tools that are incredibly good at persuasion? Is that a better world? Because in theory, you now have a deeper understanding of who that person is across from you, and a good salesperson is providing value. [00:11:55] S0: Exactly. Exactly. No. That... That's exactly how I think about it. When you can save a sales rep four hours, know, of their day, what do they do with it? I think in a lot of cases, it actually means they become a lot more consultative. Mhmm. Because it... Sales is competitive. Right? If company a uses that time to go a lot deeper with the customer and better understand their problems and come up with more creative solutions, and company b uses that time to... You know, it doesn't go... Use that time to go deeper and instead goes to more... Just goes to more customers, then who's gonna win those deals when they come head to head? So I think, like, it depends on how competitive is your space. It depends on, like, how good is your product. Like, there are a lot of factors. But I think in a lot of cases, AI making sales more efficient actually results in a much better customer experience. [00:12:42] S1: If you had to play out what this all looks like in a year from now, what do you think will be true that isn't true today?