A Product Market Fit Show | Startup Podcast for Founders
00:05:28 · Ali Khokhar
[00:05:28] S0: All the above. So multichannel, voice, text, email, and then inbound and outbound. So we're reacting, and when patients are talking to us, we're responding, or we're proactively sort of, you know, identifying certain signals that we should be reaching out to those patients. And then multimodality. Right? Voice and text. So, like, multiple channels. We could be over phone, in a web app, in a phone app, whatever the case may be, multiple modalities, and then both proactive and reactive. And so we sort of think of us as the training, the simulation engine, the orchestration engine, memory management, the underlying data platform layer, and on top of that, you build these agents across the interaction layer. So think of that as the sort of full deep stack. That's what sort of everybody was gravitating towards, because most people at that point were also like, I want to own control over how the system gets trained on what is in scope, what is out of scope. Some of the sort of clinical workflows, the clinical intelligence that's taking place, they did not wanna just buy that off the shelf, given the nature of the use cases that we had started off with. And so there was this really good fit at that moment in time where our customers were like, look, I care deeply about how this agent is trained, what the scope is, how it reasons. I want control over that. I want ownership over that. But I'm not going to build and maintain all of the underlying infrastructure to be able to sort of create that. And so it's a really good fit in terms of, you know, everybody talks about an enterprise build versus buy, build versus buy. We landed in this middle ground, which is like, you, customers, should build at the right layer, but buy a different layer, essentially.
00:06:56 · Pablo Srugo, Ali Khokhar
[00:06:56] S1: You know, you won that because of customization and control, but what would you say was the number one reason why these customers even wanted AI agents in the first place? Like, what for them was the obvious problem that this would solve? [00:07:08] S0: I mean, we know in health care there's a massive labor shortage. Right? So a lot of them had way more demand than they could handle. They could not hire and train really high quality clinical staff fast enough in order to provide the sort of very personalized hands on care that they wanted their organization to be able to deliver. And so there's this big gap that they were trying to solve for. [00:07:27] S1: And for them, is that like more revenue for them? Is that like what's the business ROI? [00:07:31] S0: It ends up being better clinical outcomes for them, so a lot of value based care organizations want to be able to drive and deliver and prove some of those things. It ends up being more revenue for certain organization or preventing revenue leakage in some cases. That's sort of where a lot of folks are focused on. Or it ends up being more revenue in the form of being able to handle, you know, two x the volume of patients without two x the volume of staff. Right? Like as we know, a lot of these organizations are very labor heavy, labor constrained, and like ops heavy. And so they're saying, okay, can I increase the demand side without having to go and find, which we know is really hard to do so, really high quality clinical staff that is trained up?
00:07:31 · Ali Khokhar, Pablo Srugo
[00:07:31] S0: It ends up being better clinical outcomes for them, so a lot of value based care organizations want to be able to drive and deliver and prove some of those things. It ends up being more revenue for certain organization or preventing revenue leakage in some cases. That's sort of where a lot of folks are focused on. Or it ends up being more revenue in the form of being able to handle, you know, two x the volume of patients without two x the volume of staff. Right? Like as we know, a lot of these organizations are very labor heavy, labor constrained, and like ops heavy. And so they're saying, okay, can I increase the demand side without having to go and find, which we know is really hard to do so, really high quality clinical staff that is trained up? [00:08:07] S1: And when was this? You you mentioned it was, like, December. What year? [00:08:10] S0: So December, January of last year, basically. [00:08:12] S1: 2425. Okay. And when did you start Amigo AI? [00:08:16] S0: The January prior. So '12 this is twelve months into our first round of sort of fundraising and capital. [00:08:21] S1: So tell me about the origin story. Now that we know that moment when things started to find click, there was a year before that where presumably it wasn't clicking yet. Right? But maybe just start with that, like, just a little bit of not not maybe your full background, but right before you started Amigo, what were you doing, and and why did you decide to start this business in the first place?
00:08:21 · Pablo Srugo, Ali Khokhar
[00:08:21] S1: So tell me about the origin story. Now that we know that moment when things started to find click, there was a year before that where presumably it wasn't clicking yet. Right? But maybe just start with that, like, just a little bit of not not maybe your full background, but right before you started Amigo, what were you doing, and and why did you decide to start this business in the first place? [00:08:37] S0: Yep. So I started my career at Google, spent time there, and then went to a company called Upwork. I was at the internal incubator there, building sort of net user to one product surface area, of like new bets for the business, new ventures. And obviously, lot of the focus Upwork, like world's largest labor marketplace, we started thinking through what does it mean to be able to deliver high quality sort of knowledge work through the form of agents. Right? So a lot of my sort of thesis and initial thinking was in that space. I left to start Amigo, which initially the thesis was similar in the sense of is there really high value labor or knowledge work that is super expensive, super hard to find, supply constrained, that we can build and train these agents for that can then go deliver that quality of work. Right? So it wasn't healthcare only in the beginning, that was the more generalized thesis.
00:08:37 · Ali Khokhar, Pablo Srugo
[00:08:37] S0: Yep. So I started my career at Google, spent time there, and then went to a company called Upwork. I was at the internal incubator there, building sort of net user to one product surface area, of like new bets for the business, new ventures. And obviously, lot of the focus Upwork, like world's largest labor marketplace, we started thinking through what does it mean to be able to deliver high quality sort of knowledge work through the form of agents. Right? So a lot of my sort of thesis and initial thinking was in that space. I left to start Amigo, which initially the thesis was similar in the sense of is there really high value labor or knowledge work that is super expensive, super hard to find, supply constrained, that we can build and train these agents for that can then go deliver that quality of work. Right? So it wasn't healthcare only in the beginning, that was the more generalized thesis. [00:09:25] S1: You know, that was the wave. I mean, end of twenty two is ChatGPT. '23 is people figuring out what AI really is, and '24 is when the wave was actually just starting, I would say, for AI agents. Is that what drove you to say, I I I gotta get in on this wave. I gotta I gotta do something with AI agents. It's too big to ignore.
00:09:25 · Pablo Srugo, Ali Khokhar
[00:09:25] S1: You know, that was the wave. I mean, end of twenty two is ChatGPT. '23 is people figuring out what AI really is, and '24 is when the wave was actually just starting, I would say, for AI agents. Is that what drove you to say, I I I gotta get in on this wave. I gotta I gotta do something with AI agents. It's too big to ignore. [00:09:39] S0: For me, it was a month after ChagiPT came out, I started raising the alarm bells saying, hey, what happens to a company like Upwork that is very labor heavy? Right? That is a labor market. That was all I was doing all day long was thinking about that problem space. Right? And so February '23 is when I actually quit my job. [00:09:58] S1: It's very interesting, like you're literally at the place where it's probably most vulnerable because it's that extra work that you don't need to hire a full time person for. You're like, throw it to Upwork, throw it to Fiverr, throw it to these places that just get it. It's the first to go.