00:00:00 · Sagar Kadakia, Brett Caughran
[00:00:00] S0: And for the first time, we could scale qualitative and start to look beyond, you know, the manual processes, right, that go into doing things like expert calls. [00:00:09] S1: We've all sort of called into a help desk, got an AI-generated voice, and sort of like it's like super, at least for me, super triggering. So can you just educate us on the curve and what's happening on voice capture? [00:00:22] S0: I can use AI to significantly improve my process. I can still have human components through that process, but ultimately my conviction is going to go up because I can do literally a 10 to 20 XN compared to what I used to. I can do way more idea generation than I used to, right? Idea generation is something net new that we have unlocked in this space because it was too complex and costly to justify to do five to 10 expert calls on every single idea. Now you can do it with qualitate. [00:00:55] S1: Welcome to the next episode of Invest with AI, where we explore the intersection of fundamental investing and artificial intelligence, or apparently we're supposed to call it super intelligence now. I'm really excited to have Sagar Kadakia join us from Qualitate AI. Qualitate is a firm that is working to reinvent the way expert network calls are done with artificial intelligence. So really excited to unpack that. Thanks for being with us, Sagar. Awesome. [00:01:22] S0: Brett, thank you so much for having me. Khe Hy, thank you so much for having me. Yeah, super excited to be here today.
00:01:22 · Sagar Kadakia, Brett Caughran
[00:01:22] S0: Brett, thank you so much for having me. Khe Hy, thank you so much for having me. Yeah, super excited to be here today. [00:01:27] S1: Maybe just be curious to sort of start with uh the genesis of the idea. Uh walk walk us through your background a little bit and and where the idea came to deploy AI into the expert network path.
00:01:42 · Sagar Kadakia
[00:01:42] S0: Yeah. Um, I've been in the primary research space for for almost 20 years and have built a bunch of predecessors uh to qualitate. Um, you know, when you think about uh primary research, there's so many different form factors, right? From multiple choice surveys all the way to expert calls and everything in between. And so that's where I've spent a lot of my career. Um and again, largely selling to uh institutions, uh, corporations, investors, um, focusing on, hey, how do we ultimately uh generate edge, right, through that form factor of primary research, right? Whether it's on the public equity side or the credit side or for product road mapping, pricing intelligence. And so have spent my entire career pretty much um uh working with those types of personas uh dealing with primary research. In terms of kind of the genesis of the idea, uh probably started back in in maybe like 2011, 2012 when I kind of jumped into the space, uh, was working um at a kind of survey-based business. We were capturing a lot of intelligence from lots of experts, you know, from cybersecurity to data management to healthcare IT. And there was just huge demand for that data uh because it was just very hard to get signals on things accelerating and decelerating. And so a lot of kind of early on in my career um got a really good sense of being able to understand the value of building out these networks, capturing information from them, and ultimately doing it in real time. And as kind of my career progressed, um, you know, a lot of the focus had been on very quantitative uh outputs, right? That had kind of been the uh form factor that people wanted for for quite some time because, you know, this was pre-LLMs. And, you know, in 2022, 2023, um, I started experimenting with a lot of the latest in voice capture tools and AI tools. You know, this is kind of when Chat GPT um started to kind of gain gain ground. And I immediately realized that um now was the time to kind of move beyond just kind of the quantitative formats of primary research. And for the first time, we could scale qualitative, right? Um and and start to look beyond the manual processes that go into doing things like expert calls, all of the back and forth on emailing and the participant selection and the scheduling and then actually having to do the calls and all the insights generation.
00:04:19 · Sagar Kadakia, Brett Caughran, Khe Hy
And because I had been familiar with this space my whole career, um, and I was leading, you know, before I started Qualitate, a lot of the uh generative AI product building, I got really comfortable using these tools and it kind of you know, the idea came to me, which was we can use a lot of the latest in voice capture technology, a lot of the latest in AI, and ultimately develop uh uh an AI interviewer that could ultimately um interview folks on behalf of individuals. And so it was kind of a journey of having built, you know, uh uh uh predecessor products to this. And then yeah, ultimately uh started the business in very early 2024, raised capital, and yeah, here we are. [00:05:05] S1: A key, a key unlock um in this business model seems to be that voice capture technology. Can you can you walk us through maybe just educate us on the on the curve? And I had an experience, you know, I think we've all sort of called into a help desk, got an AI generated voice, and sort of like it's like super, at least for me, super triggering. [00:05:26] S2: Yeah. [00:05:26] S1: But I had one the the other day where I'm like, that wasn't so bad, actually. Um so can you just educate us on the curve and what's happening on on voice capture?
00:05:36 · Sagar Kadakia
[00:05:36] S0: Yeah, I think um the reason that this business is possible is you know, somewhere in like 2022, 2023, um, like a lot of voice capture technology probably you know hit or exceeded human parity. And so one of the reasons that this business wasn't possible prior is because voice transcription was not very good. And so if you were doing thousands of um interviews using AI, you would have to have a huge team of people uh effectively cleaning up all that audio data. And I had done that before, right, in some of those companies that I had worked for earlier on in my career. And so uh because the the technology had gotten so good um at transcribing audio data, um, and we use a lot of different tools here, we've built a lot of our own orchestration on top of those tools, it was finally possible to build a company like Qualitate. And the concept of uh uh this AI interview and voice capture, you're kind of mentioning like, you know, what probably is people's uh first thought when I mentioned this, which is like, yeah, some some retailer's customer support uh hotline, right? Where they call in and those are probably all the most frustrating calls we've ever done. And so as I was starting the business, um very much the goal was this AI interviewer, um, it needs to be able to think and act and and and feel just like a human, right, like a human does. And the person on the other side, right, which is the expert or the participant, needs to have an awesome experience, or no one will ever do this. And so very early on, we started working on the technology to ensure that someone wasn't gonna have like a call center or customer service experience, right? And that we needed to um ground the AI interviewer with lots of knowledge. Um, and so that when it was having all of these conversations with folks, it could hold its own, so to speak, right? That it could ask the best possible follow-up questions, that it didn't feel generic. Um, and so those are all the things very early on that we had to solve. That and from a voice capture technology perspective, and even just like an AI interviewing perspective, we needed to build something that was just as good and frankly better than what a human could do. Or else no one is going to adopt the technology, right? Um, and so those were all the challenges that we had to solve, right? And also what we had to overcome, right?
00:08:05 · Sagar Kadakia, Brett Caughran
Because when people think about AI interviewing or AI moderation, they think about the worst call experience they've had. And so uh when we started the business, uh we we dealt with a lot of that in 2023, 2024. Um, and it was a very novel idea at that point to to do. And I think now, especially over the last you know, few few quarters, um, I think everyone is in agreement that this is where this industry is going. [00:08:31] S1: Super, super interesting. What um geez, there's so many questions. Um maybe I'll stay on the voice capture side. What um one of the one of the key things I've always liked to do my own calls, so it's like a new old dog new tricks, sort of pushing myself to learn. Yeah, is a sort of tonality, people call it body language in in in-person meetings, a sort of sentiment. Uh, even like if a if a if an interviewee um sort of gives you a answers a question with a question, is that a literal question or is that a Socratic question, right? Like those little tells often are where the real insight lies. Can you just talk to us about edge cases and sort of the interpretation of sentiment tonality, um, sort of syntax, I guess, of language and how these models intersect with those questions?