Marketers of Technology
00:03:35 · Jason Alafgani
[00:03:35] S0: Yeah, I definitely... Whenever like I'm starting a new work stream, it's a lot more of... A lot more oversight. So often, like, I'm trying to hammer it and, you know, it's almost like you start with, with marble and then you kinda chisel away to get the, you know, the finished statue that you want. And, the same is true for every one of these channels. They don't start, you know, off of the first prompt as good as they'll be. And I've... I historically thought about it as, like, sort of marble and chisel, but sometimes it feels a lot more like like hammering, like, a a square through a round hole. As soon as it feels a lot more like just hammering it, like, why aren't you getting this? But what's nice is, you try different angles, and then if it doesn't work, you can always start a new tab and try it all over again with a fresh set of context. So that... Yeah. It's usually a labor of love until you get into a good spot. And then once you're in a good spot, then you kind of scale it and you have some sort of analytics component to make sure that the things that you're putting out are getting you the output that you want. And if that starts to plateau or no longer shows the gains that you're hoping for, it can often be time to, you know, revisit it, try a new strategy. That's where I find, you know, before and after AI, the game of marketing is always trying things to a point of saturation and then needing to reevaluate and try new things.
00:04:56 · Andres Figuera, Jason Alafgani
[00:04:56] S1: Of course. No. I agree with that completely. And Jason, before I do continue asking you marketing questions and stuff, would you be able to explain to me and the audience, like, one more time, what is it that Caddi exactly does? Because I know I gave a quick intro at the beginning, but just wanna understand further how you guys work and what the service you're offering is. [00:05:13] S0: Totally. So at the highest level, it is an like an AI agents platform for operations heavy work. But we have focused primarily at law firms and at wealth management firms, both that have very often repetitive, but still complex workflows, often dealing with a lot of unstructured data, PDFs, emails, form scans, things like that. So [00:05:38] S1: I feel like those are two two pretty regulated industries. Right? [00:05:41] S0: Very regulated. And as a result, have a very high bar for accuracy. So what's cool about Caddi is it's novel approach. It builds something called hybrid agents. So they think with AI, but they act with deterministic code. So it is far more reliable and you can count on it to have RUN 4,000 be the same as RUN one. And sort of the way I view it in my head is if it's a low volume complex workflow, then something like Claude is great, you know, just reason the whole way through. But once you have to do higher volume, you don't want it to have the risk of hallucination. So in that world, it's better to try and put as much of it into deterministic running as possible.
00:05:41 · Jason Alafgani, Andres Figuera
[00:05:41] S0: Very regulated. And as a result, have a very high bar for accuracy. So what's cool about Caddi is it's novel approach. It builds something called hybrid agents. So they think with AI, but they act with deterministic code. So it is far more reliable and you can count on it to have RUN 4,000 be the same as RUN one. And sort of the way I view it in my head is if it's a low volume complex workflow, then something like Claude is great, you know, just reason the whole way through. But once you have to do higher volume, you don't want it to have the risk of hallucination. So in that world, it's better to try and put as much of it into deterministic running as possible. [00:06:25] S1: And then do clients review or or or, like, double check what the AI does just in case it hallucinates? Or with this deterministic code that you're mentioning, like they can be certain that they don't need to worry about that and that is when... What it's gonna output is what they need?
00:06:25 · Andres Figuera, Jason Alafgani
[00:06:25] S1: And then do clients review or or or, like, double check what the AI does just in case it hallucinates? Or with this deterministic code that you're mentioning, like they can be certain that they don't need to worry about that and that is when... What it's gonna output is what they need? [00:06:39] S0: Yeah. They can be almost, almost certain. There's usually like a progression to how new clients start. So when you initially build something, you will test it. So it shows you what it would have done based on the agent that you've built, what would it do if this were live data. So then you can just sort of feel good that, okay, like this would have gotten it right. Then you put it in... Into production. Sometimes people will do like a a one off run. Like there's some set of work. Let's say there are just these 100 forms that need to be processed. Let's start there and see how it does. And then they see that happen, and they're like, oh, wow. That worked perfectly. I'm happy. Or they see that there is a edge case, like, you know, the agent wasn't sure about what to do in a certain case, and then that gets flagged to a person, the person reviews it, then the person's answer then gets incorporated back into the sort of the agent's rule set. And then once people feel good about it, then they hit, you know, then they fully run, and then they stop coming to the platform. They stop checking it. We have firms that have been running for over a year that have, like, stopped logging into the platform, and work is just fully automated. So those are pretty wild. Just like
00:06:39 · Jason Alafgani, Andres Figuera
[00:06:39] S0: Yeah. They can be almost, almost certain. There's usually like a progression to how new clients start. So when you initially build something, you will test it. So it shows you what it would have done based on the agent that you've built, what would it do if this were live data. So then you can just sort of feel good that, okay, like this would have gotten it right. Then you put it in... Into production. Sometimes people will do like a a one off run. Like there's some set of work. Let's say there are just these 100 forms that need to be processed. Let's start there and see how it does. And then they see that happen, and they're like, oh, wow. That worked perfectly. I'm happy. Or they see that there is a edge case, like, you know, the agent wasn't sure about what to do in a certain case, and then that gets flagged to a person, the person reviews it, then the person's answer then gets incorporated back into the sort of the agent's rule set. And then once people feel good about it, then they hit, you know, then they fully run, and then they stop coming to the platform. They stop checking it. We have firms that have been running for over a year that have, like, stopped logging into the platform, and work is just fully automated. So those are pretty wild. Just like [00:07:49] S1: That is so cool, man. And how old is Caddi as a company? Because I know you're there since April, but how old is the company per se?
00:07:49 · Andres Figuera, Jason Alafgani
[00:07:49] S1: That is so cool, man. And how old is Caddi as a company? Because I know you're there since April, but how old is the company per se? [00:07:55] S0: Caddi is a roughly two years old, started in law firms, and over the last year has been in wealth management firms. Traditionally leaned more enterprise in both of those spaces. So Amlaw 100. So the 100 biggest law firms in the country and probably the 100 biggest RIAs in the country or investment advisers. [00:08:14] S1: Wow. And due to the regulated nature of these industries that you work with, do you only serve companies within The US or do you also work with other regions around the world? [00:08:25] S0: Just like for our own kind of business focus, we focus on The US and we focus on those two industries. But some of these companies are international, and as a result have footprints around the world. But our business focus is just, is on The US. [00:08:39] S1: Understood, understood. And now Jason, talking about more of, for example, demand generation, how are you being able to get companies to get to know Caddi, understand what you guys provide and really know the service you're offering so that they understand that they might need you guys so that you can really make their lives so much easier.