Founders in Arms
00:04:19 · Immad Akhund, Jean-Denis Greze
[00:04:19] S0: much revenue did have? [00:04:20] S2: Like, close to 1,000,000, basically, in the first year. And we were gonna do, I would say, somewhere between two and three in year two, which I think in pre AI might be okay. But you could tell how painful it was gonna be to scale the go to market. It was hard to... You really had to squint a lot to think that you could go it, like hyper go it, at least with the way we shape the business. Yeah. So, you know, you you get to this point, you're like, you stop believing that you can build something hugely successful. And for founder, you gotta look at customers and employees every day, you gotta make them believe. So as soon as you yourself no longer believe that much, it's like you feel very disingenuous. And you can be disingenuous for a little bit, but it really weighs on you over time. So we pivoted. And basically, pivoted meant we went to tell all of our customers and our partners that we couldn't serve them for next year, which is not fun. Then we had to tell people on our team that were doing tax, that were helping us build email sets for TaxRef, that they had to go and find a job elsewhere. And we did that very nicely, and we helped them find jobs. But obviously, it's very disappointing for folks that you've hired for the journey to end abruptly, even though things appear to be doing well. And then we have to go back to the dark place of of finding product market for that somewhere. I'm good, which is good. But... You know? [00:05:42] S0: So how did you come to the idea for Town? [00:05:46] S2: Yeah. Well, there's no
00:05:46 · Jean-Denis Greze, Immad Akhund
[00:05:46] S2: Yeah. Well, there's no [00:05:48] S0: Maybe maybe describe Town.
00:05:50 · Jean-Denis Greze
[00:05:50] S2: Yeah. Yeah. So Town, it's an AI system for work, meaning it is it is an AI system that tries to automate a lot of the toil that people have in their day to day, specifically knowledge workers. The insight, if you have one, would be that a lot of people, especially outside of Silicon Valley, do a lot of their work out of email and calendar. So if you can help a small business owner, a mid market salesperson, a recruiter at a thousand person company be more effective by using their email, their calendar, that's gonna make them money, and it's gonna make them look good at work, and they're gonna be willing to pay you to do that. And then they will talk about your product, and more people will use it. And so, yeah, it is a proactive assistant and really aimed at productivity around work tasks. And the way we came about it is kind of dumb, actually. So the first thing that happened is we were pursuing a few ideas, like, in the very early zero to one, trying to find design partners kind of motion, which is like a more b to b motion. And then my cofounder and I were playing with Opus three five, and we were like, hey, this thing can really fly for coding. It's much more agentic. Where else could you use this, and why doesn't no one build great AI around email? And the the second insight around email just came because when we were doing tax, even though we built really cool AI for tax rep, our tax preparers, what they were asking for was help, like, scheduling meetings, answering the same questions from their users, like being... Once a meeting is scheduled, two weeks later with a client, they wouldn't know what to talk about at the meeting. And so they were like, what if you could prepare me for the meeting and then take the action items and do a first pass at them? It was all this, like, email meeting workflow. We were like, wait. Maybe that thing that our team wanted us to to build for them, which we didn't think was as important as the actual AI tax prep. Maybe maybe now we should think about that because that seems like a pretty general problem. So we build a prototype in... I mean, the first prototype was in, like, three days, but we put it in front of users in less than two weeks, and it had very early product market fit. Meaning, like, the level of product market fit was stupid. It was...
00:08:01 · Jean-Denis Greze
There was a blank page on the Internet on our website, which is town.com. It said, we're an applied AI company and had some bullshit blog. And then the upper right, there was a sign in button that we added. And you clicked signed in, and it was like, please connect your Gmail and give us all the rights to your Gmail. Like, you know, like all the people who click people I don't know who, but I do know who after the fact. But random people that we didn't ask went to our web page, clicked on sign in, clicked on connect Gmail, looked at all the disclaimers. They're like, cool. That sounds good for me. They signed in for that. And then they saw the world's worst interface where you could basically just have an LLM rule running on top of your email. That was the initial product. So you could write a rule like, oh, if it's an email from a VC, just archive it. If it's an email from a potential lead for my business, star it and label it important and draft a reply. It was very basic. It was just, like, rules that humans could set. And for every... That would be called every email. It was quite expensive. And people just started using it, and it was like... We came back from the holiday vacation after Christmas, and it was, like, I don't know, 15 VAUs or something. We're like, 15 DAUs? Who are these people? Where are they coming from? And it kept it kept growing, and then it was, a week later, it was, like, I don't know, 30 DAUs. We're like, that's weird. What's going on? And so then my cofounder and I, we were like, everyone stop what else you're doing. We're gonna spend a month and a half building the AI on top of Gmail.
00:09:33 · Immad Akhund, Jean-Denis Greze
[00:09:33] S0: And But so it was the... It was 30 users that gave [00:09:36] S2: you enough confidence? Yeah. I I mean, if you've ever built a PLD product [00:09:40] S0: We're getting... Where are they coming from? They were telling each other about it. Is that how [00:09:44] S2: it was? Yeah. I think part of it was having town.com and having done tax. I think people were actually coming who were, like, wanted us to do tax with them from before. [00:09:53] S0: Because there's a decent number of years.
00:09:53 · Immad Akhund, Jean-Denis Greze, Rajat Suri
[00:09:53] S0: Because there's a decent number of years. [00:09:54] S2: Yeah. Yeah. Totally. People randomly showing up. Then there was VCs who signed up because VCs are always checking out what your startup is doing. So there's some of them that signed up for it. And then some of the VCs told some of their portfolio companies, and then some of the tax people told more tax people. And, yeah, it just kept on growing. In PLG, if you can get a 100 users, you can get a million users. Like, it's very... If you... No sales. PLG really means that you're not talking to anybody. You're not doing any marketing. There's... The onboarding is whatever it is. And if you can get people to sign up, retain, and word-of-mouth to get other folks, like, that is the definition of product market fit. It's very hard to do, but we got lucky into it. And so after that, we started listening to users and the product. When was this? It's like February... By... I mean, we started building, like, December 10. We started getting the first user right before Christmas. We came back from Christmas to, like, 15 to 20. A week later, we told the whole team to only build the product. We said we would... To give ourselves a month to build a better version of it. It went in private data in, like, mid February where we started, you know, tweeting about it, doing basic marketing. And by that, [00:11:05] S1: we had couple 100 users. And and how is it today? How how are we scaling from a couple 100? It's February this year, right, we're talking about? Yeah. This is this year.