00:14:18 · Sagar Kadakia, Brett Caughran
And so that's how I think about uh uh this concept of human calls and human moderation is it's very much going to uh still be a part of the process. Um, but quality, it's kind of bridging those two together. Uh, and I think it makes for an overall better, better, better, better end-to-end process. Yeah. [00:14:38] S1: I guess it's sort of our archetypal example on the tracking side too, was um, you know, for Avis and Hertz were public companies, enterprise was a private company. And so every, you know, every, you know, uh, every analyst who covered the rental cars would check in with Enterprise, you know, once once a month or twice a month to get an update on pricing. And yeah, so I would assume just for a broader net for tracking, hey, is anything changing? Um one question I have in that, like, how do you how do you ingest the baseline? Because like, how do you protect my context? Like, if I have a thesis on Avis and I want to check in with Hertz, like or check in with enterprise, it's really valuable to update my thesis, my prior calls, maybe calls with management. So the AI system can say, hey, Brett just met with Avis, CFO, two weeks ago at this conference, and the CFO said this. What do you think about that? How do you how do you build that loop while protecting that context from other clients? Because if that leaks out to other clients, then that's that's not optimal.
00:15:43 · Sagar Kadakia
[00:15:43] S0: Exactly. And so to answer your second question first, um uh all the context that our clients put into the platform, it it stays with them. Everything is walled off, right? So there's there's no training on client data, uh, there's no using client data for another interaction. We've built an AI interviewer or AI moderator that is unique to every single person, right? And so every person that we work with, every firm that we work with, right, that moderator has different context. It has our information and it has your information, and everything is really nicely kind of walled off. Um, you kind of brought up this concept of a like a what would be like a channel check as an example. And Qualitate has built the best version of that channel check. And I think this is like a great kind of use case to jump into, right? If if you're a fund. How do channel checks work? Um, every month or every quarter, your goal on a given company is to interview likely the same panel of folks with the same questions because you're ultimately trying to measure some type of demand momentum, pricing changes, new product adoption, whatever it may be. How does it work today? Uh, if you're a big firm, you probably have lots of folks calling. Uh uh, right? And and you could have 20, 30 people on a primary research team and they're doing all these calls, and that's how it works, say, very manual process. How does it work in the qualitate world? Uh, we've had a few dozen firms now move over substantial amounts of their channel checks business onto qualitate. So, how does that work? Um, you uh come on, you you pick a company, uh you just work with the agent to indicate here's all the questions uh that I have that I want to you know ask quarterly or monthly. Uh, here are uh the types of folks that I want to speak to, right? Whether it's cybersecurity experts or um you know folks in advanced manufacturing, whatever it may be, you'll build your panels, right? All through natural language, and then you will schedule that check. Uh and what will happen is every month, every three months, the AI interviewer, using all the context that you put in, will go out and do all of those checks for you. Now, here's where uh we've taken it to the next level, which is the persistent memory that the moderator has.
00:18:02 · Sagar Kadakia
So when it goes and does the second set of checks, it has knowledge of the first checks. That is now how we are. That's that's human-like. But what happens if the person on on your team leaves who's doing those checks? Well, great, now you put someone else in, they don't have prior memory, they're looking at call logs, all that stuff. The AI interviewer solves all of that. And so um persistent memory, um, context that is unique to that client, it's what makes the interview experience very unique to what you're doing. And then we have essentially been able to uh with agents and automation replicate the end-to-end workflow of that, right? We have some funds that are doing, I mean, across 15, 20 companies, you know, five to 10 checks per name, right? It's all being done through AI interviewing. And every set of checks gets better because it has the memory of the prior. And that's how you get great follow-ups, and great follow-up uh questions generate great answers from from from from those folks. And so this is an example that is one of our fastest growing use cases kind of within kind of like the you know, let's call it like the hedge fund space. Um and this is what we want to keep building on because AI interviewing or AI moderation is a point solution, but it's hey, how do you actually work that into the workflow of someone? And that's one example of how we have done it.
00:19:27 · S3
[00:19:27] S3: Sagar, in that example, can the analysts also feed in digital context to enhance the um the capabilities of the interviewer, the AI interviewer, like models or prior, you know, memos, things like that?
00:19:43 · Sagar Kadakia
[00:19:43] S0: Uh uh yes to the latter. And so any type of like uh like written context, uh transcripts, audio recordings, meeting notes, um all of that can now be uploaded right in and will ultimately um uh uh change how the moderator asks questions because the context window and the context that it has is now different. And so uh this was something that uh it took time to build because you know, things like you know, I'll use something uh very specific to funds here, like there's like a really recency curve that matters, right? Uh, which is you want to make sure you're asking uh or pulling from information that's very recent, not from six months ago or 12 months ago, because that's gonna have less relevance here. Um, and so uh yeah, like that's kind of what we wanted to build was uh an interviewer that kind of again, I mentioned earlier, like that kind of like thinks and feels and acts like you do. That has like a lot of the same knowledge that you do, right? But what's amazing about AI, obviously, is the ability for it to tap into that data in virtually real time, right? And and ask the best follow-up, right? And that's something that you know, as a human, you can't do that on every single call, right? Uh, it's just not possible. Uh, and and that's I think uh as we kind of get into now the benefits of and the superiority of AI interviewing is you know, no human could do 150 calls in a week, right? 10 checks across 15 companies, but this can, and it will be as good on the first call as it is on on the last call.
00:21:14 · Brett Caughran
[00:21:14] S1: Can we just walk through like a a worked example or an articulated example? Like say I'm looking at Nike, uh, just to pick a random name, and I'm trying to understand the curve of market share gains uh in in in their various end markets. How do I work with quality? Like, how does that so like what's the A to Z? What's the what's the system, the scheduling, the recruiting, the execution? Yeah, like what's that what's that experience look like?