A Product Market Fit Show | Startup Podcast for Founders
00:01:26 · S1, Pablo Srugo, Isaiah Granet
[00:01:26] S1: That's product market fit. Product market fit. Product market fit. I called it the product market fit question. Product market fit. Product market fit. Product market fit. Product market fit. I mean, the name of the show is Product Market Fit. [00:01:39] S2: Do you think the Product Market Fit show, has product market fit? Because if you do, then there's something you just have to do. You have to take out your phone. You have to leave the show five stars. It lets us reach more founders, and it lets us get better guests, thank you. Isaiah, welcome to the show, man. [00:01:54] S0: Thanks for having me. It's a pleasure to be here. [00:01:56] S2: So you've got this company, Bland AI. You've raised over $15 million, just did a Series B. $40 million Series B not too long ago, and as I understand it. You're doing AI for human call centers, which many people are doing but you are handling very long, like forty-five minute plus calls for companies like JPMorgan Chase. Some of these are life and death type of calls. As I understand, we'll get into that. You were telling me about the moment where you felt you were really onto something, you felt maybe this is true product market fit. You're kind of operating in this call center and there's this typhoon hits, and then what happens? I mean, take me to kind of that scene.
00:02:34 · Isaiah Granet
[00:02:34] S0: Yeah, what I always talk about with Bland is, you'll hear a lot of companies that do FAQs or what we call is phone trees that argue back to you. Our mission at Bland is we want to do the most complicated, urgent, difficult phone calls, and we had been working with this really big enterprise in the United States. Someone that most Americans would be deeply familiar with and we were basically handling a lot of their calls on the less complicated stuff as we worked our way into the more complicated stuff. And we had built all of this sort of release schedule, you know, over six months we were going to slowly handle calls where there were other things going wrong and edge cases, and other pieces. And basically one night I go to sleep, I wake up the next morning, and I've got five hundred messages from this company. They're blowing up me and their team, and they're like, I'm like, what's gone wrong? Has someone died? Has our AI gone off the grid and taken over? And no, what it had turned out was basically this big enterprise contracted a call center in the Philippines. Actually a very, very large one, and a typhoon had hit. And so basically, they decided without telling anybody to switch all of the traffic to us, to Bland. So that our AI could handle the calls with basically zero testing at that scale or the complexity and not only did we handle it. We handled it with higher resolution rate than the humans, with zero wait time, and better CSAT. So we basically, sort of in that moment, were able to go, hey, we think this is the first instance where AI has fully essentially replaced a call center. I mean, not only did we replace it, we didn't build an experience that people hate. We built something that people liked more than the current experience and that was, for me, the most rewarding possible moment of, wow, not only is this technology here, it's possible but it's possible to do this in a way that people actually enjoy.
00:04:22 · Pablo Srugo
[00:04:22] S2: Tell me more, without necessarily talking about this particular enterprise but just so it really clicks. What is happening in these calls? Partially, why are they so long? And maybe in that specific case, there's the part where it's just you're talking and I can understand how voice AI could do that maybe even better than a human with empathy and fast response, and whatever. But what about the action part? How much action is there to the things they have to do? That's where things typically trip up.
00:04:46 · Isaiah Granet
[00:04:46] S0: Yeah, that's a great question. So what's actually going on in these calls? It's really varied, and we kind of particularly play in financial services and health care, where it's really urgent. So my favorite example of a difficult call with you is a remote patient monitoring call. So this is where we call somebody, they might be ninety years old. We're walking them through putting on a blood pressure cuff, we're helping them troubleshoot that blood pressure cuff, take their reading, find it if they need to find it, plug it in, all while hearing about their grandkids, their life, any other symptoms they're having, and then taking the reading. And remember, this is on the phone, so there's no visual cue. We're taking the reading, and then we have to say, "Well, hey, they just told us a crazy number, like one hundred ninety over seventy." Are they having a hypertensive crisis, or did they put the blood pressure cuff on wrong? And how do we actually know the difference between calling an ambulance to that person's house and saying, hey, you mind just sliding it a little further up your arm or down your arm and trying that again? And these are the calls where if we don't do them, we don't do them right. It could be life or death. Similarly, we've done polls of countries where there has been recent serious unrest and we're asking questions like, are you happy with your government? And when we do these surveys, you can imagine if we were to leak the results of this, then people would be at serious risk, life and death. Those are the calls that we're talking about playing with and we don't use OpenAI, Anthropic, anybody else. So we're entirely built from the ground up. What that lets us do is be super secure and scalable. It also lets us integrate with systems really tightly. So we're reading and writing from your database, from your CRM. We're actually taking action on your behalf.
00:06:22 · Pablo Srugo, Isaiah Granet
[00:06:22] S2: But you don't use like Vapi or LiveKit or any of these things? [00:06:25] S0: No, no, we were pre-Vapi and LiveKit. Thank you very much. Yes, we are built entirely from the ground up, and that means trust also. Because there's only one provider that can go wrong, and that's us. We also don't have anyone to point the finger at, but that's why, you know, in the future, something that we want to handle is 911 calls. We want to be the first line of where you call, and we want to have that trust. And it's not just about the actions on the call, it's actually also about, on the call, we're learning live how you speak, and we're adjusting our AI speaking patterns to you using machine learning. Not just an LLM saying, what is he saying? We're actually learning from your audio patterns and figuring out how to deliver a better conversation. [00:07:05] S2: By the way, is that just a function of the fact that you started early, or would you recommend that to a voice AI company starting now? [00:07:12] S0: A voice AI company starting now should build on top of Bland but if not, I would say I would not recommend it. It was a really difficult decision. [00:07:19] S2: So you can be a piece of infrastructure, like you're not necessarily the end to end app. Silent Enterprises people can build on Bland.
00:07:24 · Isaiah Granet
[00:07:24] S0: Yeah, we have over three hundred and fifty thousand developers who are using Bland today, in addition to our a hundred and seventy plus enterprises. We do about three and a half million phone calls a week. By phone numbers, we know roughly one in four Americans has talked to Bland. Is there an element of you feeling you're running two separate businesses? Because I can understand the infrastructure pace, like, ElevenLabs or whatever. And then, actually getting enterprise and the whole enterprise sales process like there's two different worlds. What’s interesting is that we don’t think about it like that. We don’t separate infrastructure and application, and enterprise. What we really think about is we’re building phone calls, and we’re building the world’s best phone calls. And we want to make that accessible to anybody. And what the big enterprises need to be successful is a lot more handholding, a lot more in the weeds, getting direct with them, forward-deployed motion. Sort of figuring out how does this fit into the bigger picture. But that same technology can be used by anybody, really. Our job is to say, we know what a good phone call sounds like, and we refuse to allow the world to have phone trees that argue back with them. And so my short answer is, the product we build has nothing to do with the infrastructure, it has nothing to do with our application, it has everything to do with the actual call. And our real customer is the person on the other side of the phone.