A Product Market Fit Show | Startup Podcast for Founders
00:16:21 · Isaiah Granet
[00:16:21] S0: The first thing I give advice on, on pivots is it's time to pivot when you notice a couple of things. The first is, if you're no longer excited by the problem at all, pivot sooner rather than later. Sort of the advice of don't have kids if you're not sure about getting married. You know, you want to make sure that you actually are excited by the problem. The second is, is make sure that the problem you see is a problem people will pay to solve and this is really, really important. Some people like, you know, Figma, for example, can have extreme conviction that just the world will need design software and it was a brilliant vision to understand where we were going. But for the most part, what I see people make mistakes on is they say, well, I know my mom spends X, Y, Z time every day doing certain activity and this is a really painful problem, and I can solve that. It's like, great, but would your mom actually pay to solve that? And starting by leading with the money. And if we had asked deeper questions early on and said, will you pay for this? How much will you pay for this? How much will you pay to solve this problem? We would have discovered that this wasn't actually a problem people will pay to solve and that just means, unfortunately, for the time being. It's a problem that probably you can't build a business around. So start by figuring out where the money is and then when you do go into the pivot, I've seen a lot of friends have this happen. It happened to us, as I say, it's like, expect it to suck. It is going to be probably the worst, darkest days of your startup. You are going to have no direction, which for founders typically is one of the most brutal things you can experience. Because you're used to go, go, go, and your whole job is to find a problem that gets you excited and don't pick the first thing that stands out to you. Pick something that you're excited by, that you feel you have a unique advantage in, and that you can sprint at in a way that's tangible. And for us, that was two weeks of wandering and, are we gonna build software to prevent concussions? That was a very real idea that we got close to. Are we gonna build better software for getting Ubers and Lyfts? Thankfully, we didn't do that, because Waymo came out relatively soon after.
00:18:19 · Isaiah Granet, Pablo Srugo
We looked at all these ideas, and the reason why we ended up going with Bland was I was tinkering around. I guess the very real story is I didn't want to call to order food from this place that was on the street I lived on, and I was like, I bet you I can have OpenAI call it with ElevenLabs and Twilio. And, I was tinkering around and I managed to get a prototype to work. And I just had a blast. I spent like twelve hours working on it and I didn't even care. [00:18:43] S2: Twelve hours to now make a two minute phone call, but it got your business. [00:18:47] S0: Yeah, exactly. Terrible phone call by the way. It fell apart, it didn't work. But there was something about that magic and then understanding is like inherently it had value. Because if we could do these calls, if I got value out of it, then other people can get value out of it and then it turned into, well, then maybe developers, maybe businesses down, yeah. [00:19:05] S2: And then how did you think about. So you're using ElevenLabs in that thing, then you decide not to, how do you make that decision?
00:19:05 · Pablo Srugo, Isaiah Granet
[00:19:05] S2: And then how did you think about. So you're using ElevenLabs in that thing, then you decide not to, how do you make that decision? [00:19:10] S0: So post Demo Day, we are really beat down. We don't have a whole lot of money in the bank and, me and my co-founder basically are like, all right, it's time to make this thing work. We had a plan to move back into my co-founder's mom's garage in San Jose. Because we couldn't afford rent anymore and what we said was, we are not the research guys. And this was before OpenAI Realtime, before anybody came out with anything. We said, we're not the research guys. We're not going to build the best voice AI on earth, but we probably can build the best enterprise version of it. And enterprises are going to want the same things they want today, five years from now. Security, reliability, uptime, all of those core pieces and so, we're going to pull out OpenAI, Anthropic, and everybody. And just bet on this idea that if we focus on building all the core pieces in-house, we can deliver better phone calls, but we can also deliver something that Microsoft, Apple. I don't know why those were our ideal companies at the time, but those were the companies that we thought would want to buy our software if we were enterprise-oriented and so we quite literally. A shout out to Base10, our GPU provider, we put down a commitment for GPUs that was more money than we had in the bank at the time. And we bet the whole business on doing this whole going in-house. And it was a really tough decision at the time.
00:19:10 · Isaiah Granet, Pablo Srugo
[00:19:10] S0: So post Demo Day, we are really beat down. We don't have a whole lot of money in the bank and, me and my co-founder basically are like, all right, it's time to make this thing work. We had a plan to move back into my co-founder's mom's garage in San Jose. Because we couldn't afford rent anymore and what we said was, we are not the research guys. And this was before OpenAI Realtime, before anybody came out with anything. We said, we're not the research guys. We're not going to build the best voice AI on earth, but we probably can build the best enterprise version of it. And enterprises are going to want the same things they want today, five years from now. Security, reliability, uptime, all of those core pieces and so, we're going to pull out OpenAI, Anthropic, and everybody. And just bet on this idea that if we focus on building all the core pieces in-house, we can deliver better phone calls, but we can also deliver something that Microsoft, Apple. I don't know why those were our ideal companies at the time, but those were the companies that we thought would want to buy our software if we were enterprise-oriented and so we quite literally. A shout out to Base10, our GPU provider, we put down a commitment for GPUs that was more money than we had in the bank at the time. And we bet the whole business on doing this whole going in-house. And it was a really tough decision at the time. [00:20:26] S2: How do you compete with ElevenLabs? ElevenLabs is still big at that point and build your own voice AI from scratch.
00:20:26 · Pablo Srugo, Isaiah Granet
[00:20:26] S2: How do you compete with ElevenLabs? ElevenLabs is still big at that point and build your own voice AI from scratch. [00:20:32] S0: Yeah, the CTO of ElevenLabs is an investor and a friend, and I think they built an awesome product. But our job wasn't to build the best text-to-speech engine, it was to build the best engine for enterprises and we weren't focused on out-competing ElevenLabs, and we still aren't. We were focused on delivering the best possible phone calls for everybody, and so for us. What that let us do is narrow in focus and what you have to realize is early days, the distractions are so many. Your only job is to focus, focus, focus, and what people are always amazed by is not what two people can accomplish, it's what two people with focus can accomplish. And so for a founder, you should be saying no to ninety-nine point nine percent of things. That's really your job, and the zero point one percent of things that you say yes to should be your entire all-consuming life. [00:21:22] S2: How long did you have to go heads down just until you got a phone call that sounded real? [00:21:27] S0: Probably a year and a half. So if we started, it became Bland in July of 2023, give or take, rebranded. It wasn't until around October of 2024 that I said, wow, I think we just delivered a very real-sounding phone call. [00:21:42] S2: Did you raise in between or how did you make that work?
00:21:42 · Pablo Srugo, Isaiah Granet
[00:21:42] S2: Did you raise in between or how did you make that work? [00:21:46] S0: So the first problem that we identified, and this is another key part of the focus. Was we identified that this issue of latency. Which is how long does the AI take to reply to you on a phone call, was massive. It was about five seconds for us at the time in 2023. So that was from, you know, you finish talking, AI replies, five seconds and we managed to get it down to three seconds. Which was considered the fastest in the world at the time and our bet was, hey, if we can get it down to 400 milliseconds, it won't make us the best enterprise product. But it gives us an edge and a spike that we can use to drive into enterprises. And so we basically figured out how to get it down to 400 milliseconds. We did this really cool video of me talking to a phone while it's laying on the table and you can see how fast it replies. And then that ended up going mega viral. And we managed to get about $100 thousand of revenue off that ARR in January of 2024. We raised a seed round on the back of that from Upfront Ventures. [00:22:41] S2: Because the VCs already, now they like voice AI at this point.