Town vs Instinct vs GrokBot | Why the AI Assistant Market Is Not a Bubble

1. 0:00 – 2:08

   Intro

   1. JGJean-Denis "JD" Grèze0:00

      I know what I'm building is a top-three priority at Google and Apple, like, in the next 12 months. Not a top-10 priority, like a top-three priority. The product in this category that will win will have a network effect at the agent level.

   2. HSHarry Stebbings0:11

      Now, the hottest category in Silicon Valley is AI Assistants. On the consumer side, you've got Instinct. On the enterprise side, you've got Town.com, founded by today's guest, Jean-Denis, formerly CTO at Plaid. And this conversation today is probably one of the most pertinent discussions that there is.

   3. JGJean-Denis "JD" Grèze0:27

      I think you'll trust your agent to decide what data to share with other people without you intervening in five years. You can build now at the speed of machines, but you can only learn at the speed of humans. I don't think Instinct and Town are trying to do the same thing.

   4. HSHarry Stebbings0:40

      [clapperboard snaps] Ready to go? [upbeat music] JD, dude, I'm so excited for this because, in all honesty, I have a lot of founders on the show where I kind of need to pretend to be excited by their product, and I'm not really. Um, and I love Town. I was saying, the team use it here. Uh, I'm a DAU, and so I was so excited when we agreed to do this. So thank you so much for joining me today, man.

   5. JGJean-Denis "JD" Grèze1:12

      Yeah, thanks for having me, and honestly, I didn't know that you're a DAU until, like, three minutes ago, so I'm super happy. And, uh, send me all the feedback about the product, 'cause you're a DAU, but when you're a founder, you're always embarrassed about your product, at all times. So-

   6. HSHarry Stebbings1:24

      You might, you might regret saying that, but I, I will do. For those that don't know, what is Town, as specifically as possible?

   7. JGJean-Denis "JD" Grèze1:33

      [lip smacks] Uh, so we're an AI Assistant that lives in your email and your calendar, and it tries to help you do work, right? So what it looks at, it looks at things that you do already, uh, how you organize your day, emails you tend to send out, and it recommends AI automations that try to do some of the things that you would do normally yourself, uh, just do them for you in the background. And yeah, we've been a product. We've been out in the market for about three months. The, our ICP is just mainstream users, like mainstream people who, you know, use email, calendar, text messages to do work. Um, and we've had, we've been doing super well.

2. 2:08 – 4:09

   Why Town Pivoted From AI Tax to AI Assistants

   1. HSHarry Stebbings2:08

      Can I be blunt, dude? I mean, we obviously-

   2. JGJean-Denis "JD" Grèze2:10

      Yeah

   3. HSHarry Stebbings2:10

      ... did a show a couple of years ago, and I remember when you started, like, your own thing, you were doing some boring shit in finance.

   4. JGJean-Denis "JD" Grèze2:19

      [chuckles]

   5. HSHarry Stebbings2:19

      And I remember the first round going down, and I was like, "I love JD, but it's pretty boring." Um, what was the pivot?

   6. JGJean-Denis "JD" Grèze2:26

      [lip smacks] I mean, we, we, uh, so we spent a year building an AI Tax company, like business tax prep with AI, and we just... We got to some product market fit but not enough where it was gonna be a success. Like, that's, you know. There's a thing, like, the truth is we failed at building a business that would be a great business. And yeah, after a year, we were like, "We gotta reset." And we spent three months kind of in the wilderness trying to figure out what we wanted to do. And one of the, one of the areas we just looked at is why has no one built AI that operates out of email? Just there's so many people in the world who run their business and their life out of email and calendar. We were like, "Why no one? Why has no one built a great product there?" And it was just about the time that, like, Opus had gone fully agentic, right? This was, like, November, December of last year. And so it was also the same time that models could start to actually do real work, not just do, like, a few steps, but real agentic work. And the prototype, we built a quick prototype in a couple of weeks, and it had product market fit almost immediately. Um, so that was the pivot. It was very lucky. There's no, you know, we didn't go talk to, like, 100 customers and take their notes and, you know. We built for ourselves, and, um, I think it was one of those where the technology was changing so that what we wanted to do was possible at a time, um, at a time when, you know, people were very excited about trying AI products. I mean, basically, you know, OpenCall was happening, like, literally as we were building the product. Like, OpenCall was blowing up, and we were like, "Oh, shit, it's the same thing in many ways," right? They're trying to do the same thing. But it turns out there's a difference between something that's open source, and it's amazing, but it's only a kind of tinkerer that can use, and that would be OpenCall. I think with Town, we've always been focused on how do you get just about anyone to be able to get value out of the product.

3. 4:09 – 6:14

   Can Town Survive Google, Apple and OpenAI?

   1. HSHarry Stebbings4:09

      As an investor today, every company in some respects is questioned about how cannibalized could this be from any-

   2. JGJean-Denis "JD" Grèze4:17

      Hmm

   3. HSHarry Stebbings4:17

      ... of the big providers. Um-

   4. JGJean-Denis "JD" Grèze4:19

      Yeah

   5. HSHarry Stebbings4:20

      ... this is, like, right in the motherfucking sweet spot. [laughs]

   6. JGJean-Denis "JD" Grèze4:24

      Yeah.

   7. HSHarry Stebbings4:24

      Just to be blunt. Like, how do you want... And, and we've got GrokBot in the most recent-

   8. JGJean-Denis "JD" Grèze4:29

      Mm-hmm

   9. HSHarry Stebbings4:31

      ... times. How do we think about cannibalization by frontier model providers and GrokBot in recent weeks as a threat?

   10. JGJean-Denis "JD" Grèze4:38

       Do you wanna know? The, the truth is, I know what I'm building is a top-three priority at Google and Apple, like, in the next 12 months. Not a top-10 priority, like a top-three priority. So you know, I, when I go to sleep, I s- I fall asleep very quickly. It's one of my superpowers. But I do wake up at, like, 3:00 in the morning, and immediately, usually when I wake up, like, uh, some dark thought goes into there, and it's, you know, it's like roulette when you're a founder. Like, which dark thought will stop me from falling asleep again? And definitely the, uh, this is gonna be, you're, like, in the middle of the fairway and everyone's trying to, you know, is gonna try to get you. That's 100% the fear, but I gotta, I gotta say. So you wanna talk about moats, right? You're like, "What's the moat?"

   11. HSHarry Stebbings5:19

       Yeah.

   12. JGJean-Denis "JD" Grèze5:19

       "What's defensible?" I, I think ha- talking about moats is a little bit of a luxury, and you have to be more successful than Town is today for it to matter. So just, like, I, I will answer your question, but, like, my mindset right now is how do I get 100,000 or a million paying users in a market that has, like, a TAM of a billion potential, you know, paying users? Step one is, like, you need, we have to have really deep product market fit, and I think none of the products today actually have really deep product market fit yet. Like, GrokBot's really cool. It's awesome, but it's a power user product. It's not a mainstream product. Town is great, but we have a lot of work to do to make it a true mainstream product. So even before I worry about defensibility, I'm still like, "What is the right product experience that's gonna resonate with a mainstream?" 'Cause the big players are not-- They're not gonna innovate their way there. They will copy their way there, but they have to copy someone who's been successful in the first place at building a mainstream product.

4. 6:14 – 8:48

   Why AI Assistants Could Have Network Effects

   1. JGJean-Denis "JD" Grèze6:14

      Okay, then number two, like, I think the product in this category that will win will have a network effect at the agent level, right? Um, and one of the features of Town that users who figure out how to use it, they love it the most, is something called agent to agent. And that's where you ask your assistant, your townie, we call them townies. You, you ask your townie a question, and it realizes that it doesn't have the answer, but that the townie of one of your coworkers has the answer. And it just goes and asks it the question, and then that townie answers. And it's, it's-- You-- To find the feature, you have to go, like, in a sub-part of the product to use it. But that is a network effect. Once you have your whole team on that, it's actually really difficult to imagine moving to a different product. And I think no one has figured out multi-user, multi-player AI today. Like, I think that's the thing that will be the moat. In the meantime, right, uh, you-- we have lots of theories. Everyone in the market has lots of theories, right? So people will say, like, "Well, you'll post-train custom models, maybe custom models per company or per person, and that will allow you to retain your users," right? People say, "Hey, the context about a person, that's the moat, and people will be less and less willing to connect more data sources." So once you have product market fit with a user, and they really love your experience, and they've connected all the tools and all the connections, it's actually really hard for someone else to go in there because, you know, they won't have the connections, right? Other people think, three, actually, the moat-- there's no new moat in this market, and it'll be distribution. And so it's whoever already has the users who will win. So, like, I think there, for me, the, the competitor I would worry the most about would be, would be pri- for personal use cases, it would, it would be Meta and WhatsApp, right? They're, they're gonna have a personal assistant that's gonna come in WhatsApp. I don't know if they're launching it in a day or in three months, but it's coming. They already have the distribution, right? You're already-- Everyone's already using WhatsApp for, for messaging. If there's an agent in there that can do things for you, it's gonna be extremely powerful, right? So every company, right, if you, you... Some people think it's a device. Like, they think AI is changing the shape of software so that people will no longer ever go to websites. They will never use apps on their phone. The entry point for most digital interaction, much like the entry point today is either, like, a phone or a computer or a browser, the entry point will be, uh, an AI. And so whoever owns the devices is in the best place to put the AI in front of the user, and they will win. But it's... You know, this is like, I don't-- If I think like way-- If I think about all of these things, I'm like, "Oh my God, what do I do? Like, how do I, how do I win given all these competitive forces?"

   2. HSHarry Stebbings8:47

      So what is

5. 8:48 – 11:01

   Will Everyone Have One AI Agent or Many?

   1. HSHarry Stebbings8:48

      the future interaction between human and agent? And what I mean by that is, do we have, like, a consumer agent, an enterprise agent, and then a hardware agent that does maybe productivity and notes? How do we think about that, h- multi-agent versus single agent?

   2. JGJean-Denis "JD" Grèze9:05

      Each human will have one, two, maybe three entry points into the, into the digital space. Because I don't think you'll wanna be like, "Oh, I'm doing sales. Let me use the Salesforce agent. Oh, I'm doing project management. Let me use the Linear agent. Oh, I'm doing this other thing." You'll want one entry point. You won't wanna ask yourself the question. You just push a button, and you start speaking. But there are real reasons why it may be more than one, and it has to be. So one of them is just privacy and how your workplace is gonna feel about their data being intermingled with your personal data, right? So I think y- you, you might still have only one hardware entry point, but from a privacy perspective and from a, like, where your data lives, I think you're gonna always wanna separate on the data layer your personal and your work data. And in Town, we do that for you, but I think it could be two different companies that you end up using, right? Uh, like one layer, one layer below. But I think privacy is the main deter- will be the main determinant of your data silos and, and your company, whoever your workforce desire to own the data that you create for them, right? They won't want that to intermingle with your personal. But from a usability perspective, like, it's just on my phone, it's annoying that I have fifty-five apps and I'm clicking everywhere. And so if you move to a world that doesn't have hard interfaces 'cause you don't need them most of the time, you know, why, why would you have fifty agents, uh, you know, at, at the user layer. Below it's different, right? So you're an investor, I think, in, in Harvey or Lagora? Lagora, right? You're an investor in Lagora.

   3. HSHarry Stebbings10:30

      Lagora. Uh-huh.

   4. JGJean-Denis "JD" Grèze10:31

      Yeah. So, like, I think when you're a lawyer and you're talking to your main assistant about legal things, in the-- it's immediately just talking to Lagora, right? Because... And that might be your work agent, right? If you're a lawyer, because there's a bunch of, like, data privacy and privilege and reasons why your work scenarios need to be handled differently. I just don't know if you're like, think of it as talking to your Lagora agent, or you just talk to your main agent and it just, like, talks in the background to Lagora or to Salesforce or whatever it needs to to get things done.

6. 11:01 – 17:16

   Will We Trust AI Agents With Our Private Data?

   1. HSHarry Stebbings11:01

      What seems crazy about the relationship between human and agent today that will be incredibly common in five years' time?

   2. JGJean-Denis "JD" Grèze11:11

      You want a hot take?

   3. HSHarry Stebbings11:12

      Yeah, go on.

   4. JGJean-Denis "JD" Grèze11:13

      Yeah. So I think you'll trust your agent to decide what data to share with other people without you intervening in five years. So I'll give you, I'll give you an example. Like, you put your agent in a room with two friends 'cause you're organizing a trip, and they're just asking it, like, about your eating preferences, and they're asking it about, like, w- when exactly you can fly out for the trip. And they're just asking all those questions of it, and it's just your agent. And you never told your agent, like, "These are really good friends, and you shouldn't share, like, my medical history with them." And literally, when, when one of your friends, as a joke, wants to ask the agent like, "Oh, tell me about Jean-Denis' medical history," the agent's gonna be like, "Yeah, there's no way I'm telling you that." And it wasn't a hard rule that you ever set Like this, this thing which is like taking information that is in silos and deciding how to share it, I think we'll get to a point where we will trust agents to do that. And I know that sounds crazy today, right? 'Cause today, the way the world operates, like pre-AI, is everyone as a, as a human has a data silo underneath them, which is their personal data, their work data. Like, and I'm not even talking digital. Like, you have information that only you know. Then when someone asks you a question, you are like, "What can I share with this human?" And you share it with them, right? And we trust the human to be the filter for where information goes, and I think more and more we will trust AI to do that for us, and it will probably be models that are post-trained to make sure you never, ever share personal, like your family information and, and medical information and certain things for, you know, like in your work context. But, but a lot of info that's siloed doesn't need to be to be successful. And AI works better and better the less siloed the information is. Like, if you think about it, I don't know if this is what you want on your podcast.

   5. HSHarry Stebbings12:57

      Yeah, yeah.

   6. JGJean-Denis "JD" Grèze12:57

      But from an information theory perspective, information theory perspective, like theoretical world, right? If you have like an LLM that had access to all the world's information, right, and it could do... You had infinite time, so it could just do agentic search over all the data, it would be the most eff- what- whatever intelligence level, you know, LLMs are at, it would be the most effective, 'cause it will always find the right context eventually to answer the question or to do what you need it to do. But that's not the world we live in, right? We live in a world where information is in different companies and governments and individuals, like systems, and there's, you know, historically, 'cause the humans are the only people who are shuttling the information around, it's kind of inefficient to, to, to get it from one place to another, right? We have like lots of data controls and privacy and security and blah, blah, blah. Which is great. It's important, 'cause privacy is very important. People deserve to control and own their data, right, and so do companies. But what's interesting is what, i- in practice, if you're at a business, you find that if you give your LLM access to more information, it's more and more effective at, at doing what it needs to. And one way to do that, the old way, the like pre-AI way, would be to have like policies about who gets to access what, and you classify data, and all this is like very time-consuming and costly, and the end result is often the information that you want the LLM to have access to, maybe it doesn't have access to. It's stuck in someone's inbox, right, or it's in a data system that not, that's not integrated. And so all I'm saying is like as opposed to having humans in your compliance and security team over time label data and decide what goes where and what can be accessed, I think we'll just start to trust LLMs to do that, meaning you will trust your data silo and another coworker's data silos. You'll be like, "Well, I'll trust my LLM to decide what can get out of my data silo." And so then when someone else on your sales team, right, like very c- concrete example, someone on your sales team wants an intro to someone at a customer, and you're like at a 1,000-person company, and the person at the sales team knows that there must be someone at a 1,000-person company that knows the right person at that vendor, right, right, that customer. It's somewhere. And normally now they just go to Slack and they're like, "Hey, who's, who's working with client X?" Right? "Who's working with them? Who knows?" But really what they could do is their agent could g- go talk to the agents of everyone else at the company, and all those agents have access to each person's inbox, and come back and say like, "Oh, well, look, like Liz has a personal relationship with the person that you want an intro to. It's not a work one, but you could ask her if she's willing to intro. Bob has a work relationship with the person you want an intro to, and they're due to have a meeting next week. Do you just wanna see if Bob will invite you to the meeting so you have the convo?" And that's like a great business, for the company, that's a great business outcome. That's what they want to happen, right? But to do that, right, y- the individuals have to trust that it's okay for some of the information that lives in, in, in your inbox to be made available to other people at the company. And right now that seems insane. You ask me what I think in five years, I think we will be more okay with that because in practice, the LLMs will be really good at respecting privacy around things that you don't wanna share. So you don't want your salary to be shared with your coworkers. You don't want your medical history to be shared with friends. There's these things that are sacrosanct, and we get that, but you will be able to have an LLM that respects these boundaries.

   7. HSHarry Stebbings16:03

      How much wiggle room do you have on error? And what I mean by that is if you have a mistake for whatever it is, you book the wrong thing, you do, do, execute a wrong task, how much room for error do you have, and how much trust is lost?

   8. JGJean-Denis "JD" Grèze16:17

      Well, my, my claim would be like the LLMs will be much more effective at this than humans. I'm gonna tell you a story. I once worked at a place where there was a person who, uh, we'd just done an acquisition, okay? And the pr- and there was an email introducing the acquisition to the whole company, and this person had been against the acquisition. And so they meant to reply to a subset of folks, just tell them something like, "I can't believe we hired these clowns." The words may have been different than that, and instead they replied to everyone at the company. And then that person had a nickname. It was Big R, like big reply. Okay? And this, this person's like an incredible person, and they made a mistake, and it's totally fine, and everyone laughed about it, and everything was good forever after, right? But it's a human, very smart human, top .1%, who made a mistake. People make mistakes, right? Like Bob from accounting makes a mistake. Liz from, you know, HR makes the spreadsheet with people's salaries available to everyone by mistake. This happens all the time. I think the LLMs will make many fewer of these mistakes than humans pretty quickly.

7. 17:16 – 19:30

   Are Goal-Seeking AI Agents a Feature or a Bug?

   1. HSHarry Stebbings17:16

      My dearest friend is Jason Lamkin, who says that the biggest problem with agents is their goal seeking, and he talks about his agent going off and trying to buy six AP watches for him to increase culture in the company. Uh, luckily it was prevented because they needed engraving, and that was an extra step that the agent couldn't handle.

   2. JGJean-Denis "JD" Grèze17:32

      [laughs]

   3. HSHarry Stebbings17:32

      But to what extent is this maniacal goal-seeking tendency of agents a feature or a bug?

   4. JGJean-Denis "JD" Grèze17:41

      I don't have an answer for you. I think it is... I think how much you should be willing to let your agent be goal seeking and for how long you let it run autonomously is like a very, a very interesting question, right?

   5. HSHarry Stebbings17:54

      Is it your responsibility to usher people?

   6. JGJean-Denis "JD" Grèze17:58

      To usher people?

   7. HSHarry Stebbings17:59

      Guide them into, like, what is best. Hey, we find best outcomes if you let them run for X

   8. JGJean-Denis "JD" Grèze18:06

      One way to think about LLMs, right, is they're just-- you-- they turn energy into, like, GDP, right? Or into revenue. You know, like, really you step way back, right? You know, 'cause you power and, and silicon, and then you get intelligence, and we're applying the intelligence towards business results. So, you know, if they get smart enough at just creating GDP on the other end. And so, like, you just say, like, "Hey, make money for me," and then let it run for a long time, and it can do whatever it want to. Or do we wanna live in a universe where we think the human's role in this is actually to set the direction and, like, make sure that the actions that are being taken align with some kind of human value system? So I live in that second universe, where I think it, it is the human's responsibility to first allocate the resources. That means to say how many tokens are we willing to spend to try to get a goal, to set the goal, right, as well. So you, you set the goal and, and the budget. And then to monitor, right, like, the, the overall shape of the actions that are taken to get the result. And how much-

   9. HSHarry Stebbings19:08

      So-

   10. JGJean-Denis "JD" Grèze19:08

       As, as the intelligence gets smarter, you might say, well, maybe, maybe the allocating of resources, you're trusting an LLM to, like, an- analyze ahead of time what it thinks the ROI is on a poten- task and tell you how many tokens you should be willing to allocate before deciding to step away. And maybe on, like, monitoring the actions, it's also an agent that's doing that for you. I just don't think it's the same agent as the one that you put on the course to try to get to the result at the end

8. 19:30 – 22:33

   How Town Chooses Between OpenAI, Anthropic and Open Models

   1. JGJean-Denis "JD" Grèze19:30

      of the day.

   2. HSHarry Stebbings19:30

      You mentioned the different layers of kind of the value stack there. What does the model infrastructure that you sit on top of look like? How do you think about model routing for different tasks? Are you locked into one?

   3. JGJean-Denis "JD" Grèze19:44

      Yeah.

   4. HSHarry Stebbings19:44

      What does that look like?

   5. JGJean-Denis "JD" Grèze19:46

      Yeah, I, I think it's 'cause we're building an application for everyone, and we don't think most people care about understanding which model is better at what at a certain point in time, right? So we view it as like our job is to, for what you're asking for, find a model that cost effectively get-- cost effectively gets you the result that you want, right? And so, you know, for, like, concrete examples, if we're generating images, right, we have opinions internally about when we might use, like, a Gemini model or an OpenAI model to generate images, right? When we're doing voice, we have opinions about ElevenLabs when we'd use ElevenLabs to do voice, right? And, and so I think it's our job to do that because as the technology changes every day, like, right, literally every week or every two weeks there is a fundamental change, w- it's our job to make sure you get the right, the right ROI there. But it's, you know, it's, it's tough and there's, there's things that are just like, is it the right result? That, that's easy for some contexts to know what the right result is. For others, it's very open-ended. You can't know ahead of time, so you have to kind of guess, like, how difficult do I think it is and how close to the frontier do I wanna get? And then the other dimension for us is, uh, voice. People don't like it when their, their Townies sound very different, right? And one of the problems when you do model routing is there's some companies, like Anthropic spends a lot of time, I know we make fun of them online, but they spend a lot of time actually making sure all of their model families roughly don't change too much in terms of their personality. They might be slightly more verbose or less and use different phrases, but they kind of sound similar enough over time. So if you use, if you use Anthropic to generate final output for your users, it's kinda hard to suddenly move to, like, Kimi 'cause it just sounds different, so people feel like their AI's been lobotomized. So that's, like, you know... The way I think about it, for coding it matters less, interestingly, 'cause for coding you're like, "Does it work? Does the code fulfill its purpose," right? It's like, uh, yes, you might look at the code and decide it's the style that I like or not, but not really, not anymore, right? Before, when you're talking or speaking to a, to, to an assistant, if suddenly it's twice as verbose, like over text messages, it's writing you eight sentences as opposed to four, like, people don't like that. They will literally write tickets. They're like, "Despite my instructions that I gave it three weeks ago, like, my text agent seems to, you know, be capitalizing letters more," or stuff. Like, just literally you're like, "Oh, okay." So, you know, the way I think about our stack is there is the s- part of the stack that deals with the user interface, like the feel and the personality, and there it is harder for me to just route wildly because I need consistency of the experience that is sometimes hard to get from other model families. Below that-

   6. HSHarry Stebbings22:20

      Mm-hmm

   7. JGJean-Denis "JD" Grèze22:20

      ... when it's just pure reasoning and intelligence, and especially when I don't have to show as many of the traces to the users, then there, yeah, I think it's very much a matter of finding u- using the best model for the task. But, you know, we're early in this journey-

   8. HSHarry Stebbings22:32

      Can, can I ask,

9. 22:33 – 25:05

   Can AI Assistant Economics Ever Reach SaaS Margins?

   1. HSHarry Stebbings22:33

      how do you think about how a diff- how differing model providers impact ultimate economics of a user? And what I mean by that is, like, ElevenLabs is notoriously brilliant but also notoriously about, you know, the, the, the cost of a Chanel handbag. Um-

   2. JGJean-Denis "JD" Grèze22:48

      [chuckles] They are.

   3. HSHarry Stebbings22:48

      [laughs] Uh, and so my question is, how do you think about model selection balanced with cost?

   4. JGJean-Denis "JD" Grèze22:57

      Well, Harry, the answer there for every startup, uh, that I know, outside of a very few-

   5. HSHarry Stebbings23:02

      We, we don't

   6. JGJean-Denis "JD" Grèze23:03

      ... is, yeah, it's-- Well, we're hoping the price, we're hoping the cost curve a lo- makes it efficient in 18 to 24 months, right? In the meantime, you're subsidizing in part, right? Because that's what it takes to be at product market fit for these use cases because you need to use frontier for too much of the work, right? So the-- Here's how I think about it. Like, we-- So one of the things that Town does is we label emails, okay? Label- labeling emails does not in any way, shape, or form require, like, Opus-level intelligence. It does not require Sonnet-level intelligence, right? And so for that, we're already, we're already below frontier. And so-

   7. HSHarry Stebbings23:41

      Yeah

   8. JGJean-Denis "JD" Grèze23:41

      ... I think that will trend towards the cost of compute over time, right? And so I'm, I'm like, how do I use open weight models, right? Uh, how might I post-train even my own, like, smaller models? And so the, the-- it's easy to say the price of that is gonna be much smaller than it is today. And so for that part of my cogs- I don't think I stress out about it. And when I talk to other founders at AI companies, it is always so-- it's also how the thinking goes. The question that no one quite knows is how much of the workload for any particular company stays close to the frontier, where, you know, it's very expensive. And I think no- literally nobody knows the answer to that. But for human-level tasks, there's a decent amount of stuff, like scheduling movies, working with one's calendar, answering emails that have been answered before, doing research on competitors on a daily basis. Like, all these kinds of things, I think, are trending pretty far from the frontier, right? And you can already use open-weight models to do that really, really well. And so as soon as you know that, you can m- you can, you know, and you know the price every nine to twelve months halves, so you know where it's gonna be, so that you could price your product today at a point where you'll generate twenty, thirty percent margins in eighteen months. That's the way that we mostly think about it. But the open question is, at the end of the day, are you left with ten percent of your tasks being frontier or twenty percent or thirty percent? And we don't know the answer to that, and that will change the economics of, you know, these companies.

10. 25:05 – 29:56

    Open Models vs Frontier Models: Who Gets the Workloads?

    1. HSHarry Stebbings25:05

       What percent of tasks go through open versus frontier today?

    2. JGJean-Denis "JD" Grèze25:10

       Um, for us, it's mostly frontier.

    3. HSHarry Stebbings25:11

       Why is that? With the greatest of respects, the tasks being asked, I don't imagine are actually that sophisticated.

    4. JGJean-Denis "JD" Grèze25:17

       Yeah.

    5. HSHarry Stebbings25:17

       And this is where, like, sorry, with the greatest of respects, we talked about Instinct earlier. I give Instinct hard problems. Like, I want Odyssey tickets at the IMAX, continuously monitor it for days. By the minute that they're there. For you, uh, in the greatest of respects, I ask, like, for email tagging and pre-briefs. Much easier.

    6. JGJean-Denis "JD" Grèze25:36

       Yeah. Well, actually, a, a lot of people ask us to do hard things. So a lot of-

    7. HSHarry Stebbings25:40

       Oh

    8. JGJean-Denis "JD" Grèze25:40

       ... the custom workflows that people build will be quite complicated, and so that's where we will use most frontier. I think for-- We still don't use open weight for things like labeling, but we will use much cheaper models from one of the frontier providers. The, the main reason is because as a company, our focus... Like, imagine I can improve our cogs by moving to open weight, but it doesn't give me much product advantage. It doesn't make my product work better. So if I have an engineer hour, what is the engineer hour best spent on? Is it taking my current AR and making it more efficient, or is it figuring out a way to grow the product faster by working on a better network effect feature or making the model better and integrating with a new data source that makes its trajectories much better for a set of our users? And we're very focused on growing the pie faster, much more so than getting the ideal economics. And our economics are fine, right? They could be better, right? And I could move down the cost curve faster, but that's not the constraint to success for the business, so that's why we don't do it.

    9. HSHarry Stebbings26:45

       On the network effect side, have there been any interesting lessons or observations? What have you learned on expansion from wall to wall?

    10. JGJean-Denis "JD" Grèze26:53

        So first of all, interesting aspect for us of network effects is if the company has one person who is a tinkerer and who starts to build things like team skills, team integrations, team routines, that is like building-- those are all building blocks on Town that everyone on the team gets for free, then we tend to see a lot more adoption faster. And it- it's, it's interesting, right? 'Cause I'm-- we're trying to build a product that doesn't require the tinkerer, right? Because for the single-player experience, we want you to be onboarded and get a ton of value, right? And if you're a real estate agent, you have automations that are real estate agent-specific, and if you're a salesperson, you get automations that are salesperson-specific. We're trying to give that, that experience to you out of the box. But what happens is, if there's a power user that's next to other users, they find ways to make those other users much more successful. And I think you'll find that with a lot of AI products, right? So one of the, one of the predictors actually is, is there a tinkerer on the team? So one of the questions we ask ourself a lot is, "Can we identify those folks, and can we make it easier for them to create virality for their other team members?" Um, and then the second observation is there are a lot of functions that are underserved by AI within companies, even enterprises. So I'll give you a very simple one. If you're a sales team at an enterprise, you've been sold AI, like, in every direction now for three years. Do you know what I mean? Like, literally, like, if you're a sales ops person and you do not have ten emails a day from, like, an AI company, you've, you're something... You've not on LinkedIn. Like, something's wrong.

    11. HSHarry Stebbings28:21

        Totally.

    12. JGJean-Denis "JD" Grèze28:21

        But then there are, there are other functions like, uh, executive assistants, chiefs of staff, uh, uh, HR team members, finance, like some, some more junior finance team members that don't, there's, don't have that much AI in their day-to-day. They, like, really don't. And a lot of their workflows still operate out of email or, like, recruiters, right? And so a product like Town, I think often we find early adoption and growth, and I would say, like, some functions that seem from the outside like less juicy maybe, but actually that are really hungry for technology to help make their life easier. And so as soon as they adopt it, there's an interesting effect because they often work with, like, leaders or execs or people in ops teams, and then we get penetration through the op teams.

    13. HSHarry Stebbings29:06

        How important is time to wow or time to user delight?

    14. JGJean-Denis "JD" Grèze29:11

        I mean, yeah, I think the only reason our product works today, honestly, is we have very low in, very low time to value for a single user. And so what happens is single users like the product, even for sim- You were talking about, like, meeting briefings and kind of naysaying, but coming into a meeting really fucking prepared, getting the action items from the meetings, like, automatically handled for you is, for a lot of, for a lot of roles, like sales, recruiting, small business owners, is like magic. So the fact that with us, no configuration, you get out of the box, it gives a magic moment. Then the user is, like, asking questions. "What else could I do with this technology?" That's for us all the... Like, our focus is on getting that right. Can we give you time to value there really quickly? And then we play the longer game on all the other integrations. And for us, that's working really, really well.

11. 29:56 – 33:45

    How Should VCs Invest in the AI Assistant Race?

    1. HSHarry Stebbings29:56

       Since we've had Town, and again, sorry, but Instinct in the last two to three weeks break out seemingly so- I've been pitched four to five European Towns or European Instincts. I mean, literally four to five separate ones. Some are work enterprise, some were consumer, some were both. How should investors be thinking about this space, if you could advise us?

    2. JGJean-Denis "JD" Grèze30:20

       We were talking about moats, you know, earlier. So, you know, we're making... Like at Town, we make some bets, right, on why we think the product will last. I'll answer your question about Europeans, but we've made some bets. So like one of the bets that we made is you only get one AI Assistant, it has a name, you give it an image, we call it a Townie, and we have a whole brand around really building this relationship between the human and the AI. And for a lot of our users, that resonates. Like, they want that. They want something like that they trust and, uh, that they shape and that they color, that's in their image, that they name. They like that a lot. And it's silly, but actually I think that is a form of defensibility. Much like Snapchat structurally in the market is defensible even though it's not nearly as good a business as TikTok or as Facebook or Instagram. Because it's fundamentally different, it has a strong opinion about how it operates, there's some people that are drawn to that opinion. So we have a very strong opinion about the relationship between the user and their Townie, and for a lot of our users, that, that really resonates. So we have that. We have a bet on network effects, which we've talked a whole bunch about. And then the third bet that we make is we are big believers that through a lot of pre-processing, you can get better outcomes for users. So we spend a lot of compute before you even ask a question to build a mental model of the user, like pre-c- pre-creating context in a way that allows us to be really effective at things like, uh, n-n... like work networking, right? Understanding the projects that you're working on, understanding your company and, and, you know, those kinds of building blocks. So those are like three things that we think over time make a difference for a product. So the question, if I were an investor, is like why does someone deserve to win in Europe? And is it a distribution thing? Is it a GDPR and privacy thing? Is it a like Town is not, in fact, doing any marketing in France, and so you could win in France if you're the Town of France. And so you have to have some specific reason why s- some local company will win endgame, 'cause this is... These products are expensive to build. Like you said GrokBot earlier. My R&D, a lot of my R&D is just keeping up with the Joneses. Do you know? It's just like we need to just do what you have to do. Your agent has to be as capable at least, right? It... Forget your distribution strategy, forget all, like, the fact that you're good at work, forget, like, the network effect features, but if Codex can do something that you cannot do, and that thing is something that matters to users, it's over, right? So you always have to be, like, at least as good as a harness and capabilities as everyone. It's very expensive to do that, right? It's not like I have two engineers of the team trying to keep up with Codex, right? Codex is like 100 people making that thing better, and so we have to somehow, right, be as effective on a lot of tasks as Codex, otherwise the user's gonna be like, "Why would I pay $50 a month for Town? It doesn't make sense. I'm gonna pay $24.99 for OpenAI." Right? And so I think that is... If I were an investor, I would be like, "Is this local competitor have enough TAM? Are they gonna be able to have a war chest that's big enough to just keep up with capabilities? And then do I believe they have a, a reason to win in this very horizontal market like where they are?" It'll be hard.

    3. HSHarry Stebbings33:28

       Is hiring in the Valley as insane as everyone says it is?

    4. JGJean-Denis "JD" Grèze33:31

       So I don't find it crazy, honestly. Like, I think when I was at Plaid and we were competing with talent for, like, Stripe, that felt like no harder than what I'm doing now.

    5. HSHarry Stebbings33:40

       Can I ask what's been the hardest thing about the product build that you maybe didn't expect?

12. 33:45 – 39:31

    Why AI Startups Can No Longer Outrun Their Competitors

    1. JGJean-Denis "JD" Grèze33:45

       The speed of the market is insane, Harry. I've never seen anything like it. Before, in the, in the past, you know, you, you, you talk to customers, you build a feature, they would use it, and you would be, like, learning from that feature, and the learnings would go into the next feature and the next feature, and then eventually someone would copy your first feature. But then you had like th- you were three learnings ahead. Do you know what I mean? You'd been able to, like, use your product market fit to generate more product market fit. And, you know, for startups, usually you're, like, milking these, these, these user insights for a really long time. Eventually, you run out of user insights, but that would take like 10 years, right, to happen. So you have this entire period where you can, you can just because you're number one or number two in a market, learn more and iterate, right? And it's really, it's really good. The problem today is it is so much faster to build that as soon as something is working for somebody, everyone notices and is able to get there within, like, two weeks or four weeks, like copy really, really fast. And learn- You can only learn at the speed of humans. Do you know what I mean? You can build now at the speed of machines, but you can only learn at the speed of humans. And so you're... I'm not able to extract quite as many learnings as, you know, s- that allows me to the, get to the next feature. So the way it feels right now is, like, speed is such a... It's so necessary, but everyone's moving fast, right? And, and, you know, when we started the company, like, I would think first quarter and second quarter of this year, my mentality is like there's like 15 competitors in the startup universe that are competing with us, like maybe 15 companies that matter. And now I'm probably down to, like, two or three competitors. Like, I feel like, you know, it's mostly, it's, it's gonna be, like, a f- only a couple companies are gonna win this space, right? And we're, like, close, like from the startup, from the startup starting gate. There is the, th- there's the, like, Apple, Google, SpaceX Grok, uh, GrokBot cursor, you know, uh, OpenAI and Anthropic starting gates. Like, I'm s- But you have to at least, you have to clear the clouds, right, for the startups. And usually also the startups would be really fast, but the established players wouldn't be that fast. But you gotta be honest, like Claude, like Anthropic is very fast. Cursor, GrokBot, they're operating like a, at a speed that is, it's uncanny, you know? And one of my best friends runs engineering over there, so I'm always like, you know, whenever I [chuckles] talk to him, I'm, I'm sad that we're competing. You know, we're competing and I'm like, "That guy's good." Like, that dude can get shit done. I gotta beat one of the best people in the Valley at a company that has the DNA of a startup but is operating with huge cost and scale advantages, right? So, you know, that's what's super stressful because I think you can't rest for one minute. You don't have the feeling that the competitors, it'll take them six to 12 months to catch up. Um, and I feel that, like, never, yeah, never before. But it's also the most fun time to build, so you know what? You know, you can complain-

    2. HSHarry Stebbings36:36

       JD, can I-

    3. JGJean-Denis "JD" Grèze36:37

       But it's fun

    4. HSHarry Stebbings36:37

       ... you, you mentioned, you mentioned the two to three that matter on the startup side. Who would you say those are and why'd you choose them?

    5. JGJean-Denis "JD" Grèze36:45

       No, I'm not gonna say that.

    6. HSHarry Stebbings36:47

       [laughs]

    7. JGJean-Denis "JD" Grèze36:47

       I'm not gonna give free marketing. I'm not gonna give free marketing to competitors.

    8. HSHarry Stebbings36:52

       Oh, dear. You, you've got... You can't blame me for trying.

    9. JGJean-Denis "JD" Grèze36:55

       No, no. It's, it's-

    10. HSHarry Stebbings36:55

        I tried to do the Louis Theroux. You know, it's like, "Hey, how do you think about that? Like, tell me." [laughs]

    11. JGJean-Denis "JD" Grèze37:02

        Yeah. I mean, you know-

    12. HSHarry Stebbings37:04

        [laughs]

    13. JGJean-Denis "JD" Grèze37:06

        ... we... I said it before. Look, it's a blue ocean market, right? You gotta understand. Like, I never... When we go to a, to most customers, they've not heard of anything. It's blue ocean 'cause people, people are u- u- using ChatGPT as a, like, a Google enhancer. Like, that's the market. So, you know, competitors, competitors are, are great. They put pressure on you. They make you feel like you have to execute at a really high level. But what's important is you have a different strategy than a potential competitor. Like, you, you mentioned Instinct earlier. Like, I don't think Instinct and Town are trying to do the same thing.

    14. HSHarry Stebbings37:37

        You don't?

    15. JGJean-Denis "JD" Grèze37:38

        I don't think we're trying to monetize in the same way. I don't think so. I mean, w- we will see endgame, but I think what we do and generate... We generate revenue, right, from companies, right, that are using us for work with network effects around multi- multiple team members working on it. Like, their product doesn't do any of that. Maybe that is part of their strategy. I see a strategy that's more like customer acquisition, like, with a free product that's fully subsidized right now. That might change, right? But if you look from the outside, the products have similar capabilities, but all harnesses have similar capabilities. But if you look at the ICPs, where the marketing is going, it just feels pretty different to me. So I pay attention to, like, a GrokBot more than I would to Instinct 'cause I think GrokBot is going after a similar market to what we are, right? And so that is, for me, that is more the place where I'm like how is our strategy differentiated from GrokBot? How are we gonna acquire a different customer? How are our capabilities and our harness gonna really stand out and feel really different to users? Like, that's more, like, my mindset than, you know, at least-

    16. HSHarry Stebbings38:36

        To what extent is GrokBot's integration into X a feature or a bug? 'Cause to, to some corporates into professional usage, it could be concerning, actually-

    17. JGJean-Denis "JD" Grèze38:47

        Yeah, I think-

    18. HSHarry Stebbings38:47

        ... the integration with the search

    19. JGJean-Denis "JD" Grèze38:48

        ... well, I think they have a, I think, brand for them. I think some people just won't wanna touch it because of brand, and it's like that's just inevitable and it's, you know, that's just a thing that they're gonna have to deal with forever. Um, but from a distribution perspective, uh, and I think for some segments, I think in early growth it is probably quite useful. I think a person on X that uses these products is not actually product-market fit, meaning that those are not the mainstream users and ev- You have to keep that in mind. And I, I, by the way, I think they think about that over at, at, at Gr- GrokBot all the time, right? I don't think they think winning the power user/influencer on X is where the market is. That is not where you win the market. That's the early

13. 39:31 – 41:50

    Why Apple Could Lose the AI Assistant Race

    1. JGJean-Denis "JD" Grèze39:31

       adopter market, but...

    2. HSHarry Stebbings39:32

       Can you help me understand Apple's agent roadmap?

    3. JGJean-Denis "JD" Grèze39:36

       I think the problem for Apple is twofold. One, they're not a cloud company. They're just not. It's just not their DNA. They don't know how to do cloud. And the reason why that matters is because what we talked about earlier, agents are better the more data they have, and the data is not all on the phone. And, and so the fact that they're not cloud is, like, one big issue. The second issue is they've, like, contorted themselves for competitive reasons around a privacy and on-device story that is, like, absolutely f- like, puts them far away from the frontier. Like, local models on the phone, it's amazing, but they're just not, they're just... It's just slower and dumber, right, than what's at the frontier. And so as long as they're committed to this, like, on-device, like, privacy-preserving stuff, it's... The privacy stance is good from a human perspective, but they've tied it too much to the on-device. So the la- not being good at cloud, then being on-device, and then from the privacy standpoint, making it difficult even on the phone to interoperate with all the data that they have, it just, it's... Those are a lot of disadvantages to play with. Now, you know, at the same time, they do have the devices, so the new Siri, right, is gonna be mu- a much better personal assistant than the... I mean, it's rumored, but you know. You know people who've tried it, so when you're in the Valley you've, you've... You know? And so we all know it's gonna be good. But I think it's gonna be good, but it's gonna feel ne- not nearly as powerful as Town or GrokBot. Like, it's not even gonna be there in terms of its capabilities. But it'll be on your phone, it'll be convenient, you'll be en- uh, able to, like, enable more data with it. It'll have a cloud component. It's gonna be good, but it's gonna be like nine months away, I think, capability-wise compared to, you know, uh, everything that we've been talking about today. And I just... I don't know. I mean, they have a new CEO, and we'll see how they take it, but I think they just need to hire somebody that has totally different DNA and be like, "You guys, you don't understand. Like, the way people interface with digital data is changing, and we either are figuring this out, and we may have to throw a lot of our principles away, or we're just not gonna win this generation of the war." And I th- I think that is, like, a, a real risk for them.

14. 41:50 – 43:42

    Are AI Agents Creating a Cybersecurity Time Bomb?

    1. HSHarry Stebbings41:50

       Are you concerned by the data leakages that we're gonna have and the, the kind of golden age of cyber threats that we're entering into? It seems like we've all kind of normalized cyber attacks, and it's like, "Ah, Meta had one. Ah, Anthropic had one."

    2. JGJean-Denis "JD" Grèze42:02

       If you wanna build AI that is used in, for business use cases- You, you must, you cannot get it wrong. We've passed the point where humans will read every line of code. That is never happening again. The most important lines of code around access controls and things like that for systems are still being read by humans. But overall, in the history of humanity, we have passed the point where we will go back to a world where humans are looking at lines of code. It's computers that are building code that is being shipped into production with various guardrails from testing to other models, like friendly models attacking you so that unfriendly models can't later find exploitations. That's the world we live in. Obviously, in the history of humanity, in this new world, there's gonna be points where there's bad events that happen. That's just... You know, it's a little bit like chemicals, the way I think about it. Like, you know, in, like the 20th century, we started to do cool things with chemicals, and then we would put the chemicals in rivers. And then cities downstream, people got sick. And then we were like, "Oh, yeah, okay. Let's pass, like regulation like the EPA so that you can't just dump the chemicals in the river. You gotta, like clean them a little bit before you do." And then later you, like label dangerous chemicals not as dangerous, where they can go, how you get rid of them. Like, we learned along the way. There's a set of best practices, both, like from a regulatory perspective and just, like best practices in industry. And right now, what's happened is the, like cost-benefit of attacks is just thrown out of whack, and we're trying to figure out what the best practices look like. And you can't imagine we're gonna get that right every step of the way, but I think we will have to because there's no way we're going back to a world where humans are looking at every line of code.

15. 43:42 – 47:46

    Why Token Maxing Is the Wrong AI Metric

    1. HSHarry Stebbings43:42

       Can I ask you, when we think about usage, how do you define a successful user?

    2. JGJean-Denis "JD" Grèze43:47

       Oh. Um, well, I, I just define them as someone who pays me every month. If they keep paying me-

    3. HSHarry Stebbings43:54

       [laughs]

    4. JGJean-Denis "JD" Grèze43:54

       No, I'm, I'm serious. I'm serious, right? If they keep paying me, I've done my job, right? I can't think of them as the most, the more tokens they use. It's a dangerous way to think about it. Because if you think about it as, like they use more tokens every month, that's successful, what if they're using the tokens in a way where the ROI is less clear to them? Meaning-

    5. HSHarry Stebbings44:14

       Mm

    6. JGJean-Denis "JD" Grèze44:14

       ... they don't realize that they're using tokens to do things that they don't value as much. If you do too much of that, then they wake up one day and they're just paying you too much, and they get mad, and they churn off of the product. So I think you have to take a long-term perspective. The, the problem with token maxing, there's two problems in my opinion. One is like companies told them, told people, "Hey, you can use as much money as you want on AI," which is bad. Like, you want people to think, "Is the ROI of using AI here worthwhile?" So now you know, people are not going against e- employees using AI. They're going... They're like, "No, no, we need incentives so they use AI for good reason." So that's like one aspect you need to think about ROI up front. But the second problem is sometimes it's hard to know the ROI of something like, let me be more prepared for a, a meeting. How mu- how worth it is it to be more prepared for a meeting? For people who have back-to-back meetings all day, not having, being able to in, like one minute before the meeting feel prepared enough to not look like an idiot, it might be worth quite a lot. For people who have, like meetings where they have someone else preparing and/or presenting in the meeting and they don't have to present anything, and they're just sitting there, it's not worth anything, right? So that... I'm just using this example, like this workflow has totally different value for different folks, but it costs the exact same number of tokens, right? And I don't think that people think about it that way. So I think of success as paying me, 'cause if you're paying me every month, that means I'm mostly doing a job of delivering enough value when you look at the $49 or the $15 or the $99, whichever plan you're on on Town that you pay me, you're getting enough value. But I'm very concerned along the way with informing you about where you're spending money because I think you need to feel like I am doing a good job of avoiding you spending too many tokens. So one of the most popular features for us for the last month is we started sending emails when it looked like you had, like rogue routines, like routines that were just co- costing a lot of tokens. And then people were like, "Oh, thank you. I trust you. Now I feel that you're, you're looking out for me using the product badly." And that's the part for me, I don't know how to measure it, but I want people to... Success for me is you pay me and you trust that we're the right platform that helps you both use AI but do so efficiently. And I think if we can do that, we can, we can have a pretty decent business.

    7. HSHarry Stebbings46:25

       Yours is $14 a month, $49 a month, and $99 a month?

    8. JGJean-Denis "JD" Grèze46:30

       Yeah, and $199, correct. But the $15 plan-

    9. HSHarry Stebbings46:32

       Which are, whi- whi- which i- which is the most profitable segment, and which is the least profitable segment? And the reason I think about that is, like my friend Jason Lemkin, he obviously pays for, like Anthropic Pro or whatever it is, $299, and he spends about $15,000 of tokens. He is the worst customer for Anthropic, but he's on that, like Pro Max individual plan.

    10. JGJean-Denis "JD" Grèze46:53

        Yeah. We don't, we don't have a Max plan, and we have users who ask for it. And I've been asking myself, like, "Do we let the whales have a Max plan?" Because from a marketing perspective, it's useful. You know, they're just advocating for the product all the time. So I've thought about that. Um, [ts] yeah, the, the $15 plan is a really good deal, I would say, for users. It's mostly... I would say it's like a, we use it as a way for people to use the product enough that they realize they should p- pay $49, where the product is really powerful. So $15 has the worst, the $15 plan has the worst unit economics, is the most subsidized. And then I would say probably the $99 plan is the most profitable overall because it's like a power user, but it's not a power user that's, like trying to, like, y- you know, spend unlimited numbers of spend. But we also have usage-based pricing, right? So what happens for us is once you run into the plan limits, mostly you go to usage based. And so, right? Like the, the... We, we try to adapt the spend

16. 47:46 – 50:06

    Consumer vs Enterprise AI: Where Does the Bigger Business Get Built?

    1. JGJean-Denis "JD" Grèze47:46

       to the user.

    2. HSHarry Stebbings47:46

       There's... Can you choose one outcome for me? 100 million consumers paying 20 bucks per month or a million customers paying 100 bucks a month?

    3. JGJean-Denis "JD" Grèze47:58

       No, the more, more users. More users paying less.

    4. HSHarry Stebbings48:04

       Why is that?

    5. JGJean-Denis "JD" Grèze48:04

       Because I think over time, in, in work, in the work setting, AI will- Be used to do more and more and more for people. So I think the long-term potential for growing, for like NR, driving more revenue per user is extremely large over the long term. So you want to acquire the users in a paying motion because you want them to be for work use cases because you will keep finding more ways for them to use AI to generate business value for themselves. Whereas in the personal sphere, it doesn't feel like that to me. Like, you know, I only have so many, like, restaur-restaurant dates I need to book with my wife or, or trips I need to organize with my friends. I only have... Like, I only, I only have so many, like, personal, like, doctor's files that I need to send to a new doctor. Like, there's only so many of those things, and when I do those things, I save time, and time is worth money. But on the business side, when I create something that generates value for the business, they make more money, and then they want more of that thing. So, like, like, a clear example for you, if you want to... Like, r- a recruiting firm on the platform, they can take more clients because of Town. They've used Town to automate enough of the recruiting process that they literally take more clients. And for them, taking an incremental, and without hiring anyone, you, at, uh, at this recruiting company, one more client is, like, an extra $3,000 a month, right? And they pay, they pay us, like, you know, across all their users, like, five, $600 a month. And so the, like, ROI is, like, super simple for them for a business use case. They're like, "Oh, I pay $600 a month, and I get $3,000 of revenue." And that is like, that makes sense. I like that. They're making more money. Everyone's happy. And I think for them, if I could show them the way that they could take another client, and even if it costs them another $500 on Town, they would be willing to do that. And so I think the elasti- like, the, the growth potential on the, on the business side is much larger. So I'd rather have lots of users paying us less because I think over time I can show them that I can deliver more and more value, and it's worth it for them to spend more and more on Town.

    6. HSHarry Stebbings49:57

       What is Town not able to do because of model capability that you think will be incredible in two to three years?

17. 50:06 – 51:57

    Is ElevenLabs Worth $22BN?

    1. JGJean-Denis "JD" Grèze50:06

       I mean, voice. Uh, voice is so obvious. It's ha- I mean, it's happening right now. But, but it's not voice like you just speak to it. I just mean conversational. Like, I, like-

    2. HSHarry Stebbings50:16

       Would you be-

    3. JGJean-Denis "JD" Grèze50:16

       You know-

    4. HSHarry Stebbings50:16

       Would you be an investor in ElevenLabs at a $22 billion price?

    5. JGJean-Denis "JD" Grèze50:20

       I think on ElevenLabs, like, we're users of ElevenLabs. I mean, I just think it's just they sound the best. I'm not paid. I'm not an investor. It is very expensive. What I don't know is if it tops out, and that's, I think, the risk for something like ElevenLabs, meaning, like, we just get voice is good enough, and then you can get it... I can put, you know, op- open weight models on BaseTen and get it. But it just doesn't feel like that right now. I just don't know how much runway they have before it reaches that, and so that's why I'm not saying I'm bearish. I really like that company. But, like, $22 billion's a lot of money and, um-

    6. HSHarry Stebbings50:50

       And how price sensitive are you in a year or two? With the greatest of respect, right now, you can, you can burn cash. It's about PMF and growth and-

    7. JGJean-Denis "JD" Grèze50:58

       Correct

    8. HSHarry Stebbings50:58

       ... beating others. In a two to three-year, where you're bluntly trying to make economics work in a much more efficient manner, ah, if we have two to three million users using voice, dude, that's, that's a hit to our margin profile.

    9. JGJean-Denis "JD" Grèze51:13

       For sure. I mean, I care a lot more about that time. That's why I'm saying, like, if the voice capability... It'd have to be, like, maybe twice as good as today, especially tone and expression and, like, the emotional read. Like, once you solve that, I would want to go as cheap as possible because especially for, like... O- once it feels good enough, it's almost there. Like, I don't need much better. But, you know, on the margin profile stuff, I don't ever think of it as burning money. I don't... That's like my mom, my parents would not be okay with me saying words like that. So I think we are being thoughtful in our spend in order to, you know, optimize for growth in the short term and gross margin i- in the long term. The voice is not, though, where I'm really stressed out about it, honestly. But the, you know, there is a question-

    10. HSHarry Stebbings51:55

        What are you really stre- Why are you really stressed

18. 51:57 – 1:00:01

    The Biggest Risk to AI Assistant Margins

    1. HSHarry Stebbings51:57

       out about it?

    2. JGJean-Denis "JD" Grèze51:57

       It's just I think it's the percentage of tasks that are frontier. Because on everything else, I can imagine getting the prices down. But, you know, the thesis that I just said before is over time, there are more ways to use AI to generate more revenue for a lot of companies. The implication there is, is like, there's, like, things at the frontier that generate more revenue, and the problem with the frontier is I have zero pricing power at the frontier. And I mean, I think this is what happened to Cursor, right, at the end is, like, you can have huge market share and be... The customers love you and everything, but if you're, if you're paying your suppliers and competing with your suppliers at 70% margin, eventually it gets, like, a little bit difficult. And so, you know, the, that, that is the part that I'm worried about, the end game. If we are... But again, I have lots of ifs. I have to get to tens of millions of users. They have to be paying. I have to have a lot of scale. And then I'm, like, at a place where I'm still competing with my suppliers. I'm still competing with OpenAI and Anthropic, and I'm just giving them money for the 20 or 30% of workloads that are at the frontier for me, and that's what makes the economics not work. And that's the, that's the part where at the end of the game, I need some, I need some solve for that by then. I don't need it right now. The reason that's the only problem is that's the only part of my economics that's different from somebody else's. Okay? So then there's the macro question, which is, is all AI... You know, is there not real product market fit for AI products because it's all subsidized, right? That would be, like, the other take that some people could have. But otherwise, as long as you're not competing with your suppliers, you have the same economics as your competitors, and so then your ability to drive margin usually is driven by the competitive landscape more so than anything else, right? And so the fewer competitors you have, the more margin you can have.

    3. HSHarry Stebbings53:28

       You are competing with your suppliers. I mean, like, Astra-

    4. JGJean-Denis "JD" Grèze53:31

       I am

    5. HSHarry Stebbings53:31

       ... you see as, you see as a direct competitor, correct?

    6. JGJean-Denis "JD" Grèze53:33

       Totally. I am today. 100% I am today. Yeah. But the percentage, the percentage... That was a face. The per- I mean, I am competing with Astra, but there's still the, the... It's the open box problem. You would be shocked at how many people just don't know what AI can do, right? It's like the problem is having a product experience that gets a normal person to get value out of AI is really hard. That's why when people use Town, they like it, and then they start paying for it. Like, the, the, the, the payment rate for us on acquisition is, like, more than 15% of users who try the product end up paying for it, which is extremely high for PLG. Because the value delivered relative to what they were getting out of ChatGPT is just huge. Yes, I'm competing with-

    7. HSHarry Stebbings54:18

       Why? What, what did you crack that other people didn't to get that 15%?

    8. JGJean-Denis "JD" Grèze54:23

       It's ... The only, the really insight behind the product was if you ask people upfront to connect their email and their calendar, you can know enough about them that you can suggest tasks that AI can do for them. That's, like, the only insight. And i- if you're a chat- if you're working at OpenAI and your chat-

    9. HSHarry Stebbings54:41

       Is it ... I'm not being rude. I'm sorry. This is where we joked before about me being more mouthy and gobby.

    10. JGJean-Denis "JD" Grèze54:45

        Do it. Do it. Go on.

    11. HSHarry Stebbings54:45

        Is that that in- is that that insightful? Like, ChatGPT's always, like, going, "Here's all the things that we want from you," and I'm like, "No way, no way. [laughs] Read, write email abilities, no." [laughs]

    12. JGJean-Denis "JD" Grèze54:56

        I think, well, I think their suggestions are, I think their suggestions are just, like, plain bad, to be honest, but they didn't even have suggestions until a few months ago. The delta is this. If you want to use ChatGPT, you don't have to connect your email. They just don't force you to do it. They ask you a bunch of times to do it now 'cause they realize the value's helpful, but the base experience, they're trying to show someone normal, "Hey, you can have value in this product just 'cause you have a chat box." And that's how most people experience it. Our approach is more like, "Listen, you have to connect email and calendar. You cannot use our product if you don't do those things. But if you do those things, we can do all this magic for you, right? Here's what we know about you. Here's work that you normally do. We'll recommend automations that automate that part of your work." That's the, that's the part where people are like, "Oh, that's really cool."

    13. HSHarry Stebbings55:38

        I think of it a bit like a, a hard paywall, you know, when you land and it's like, "Hey, pay your monthly subscription." You're like, "Hey, connect your calendar and your, your email."

    14. JGJean-Denis "JD" Grèze55:47

        Yeah.

    15. HSHarry Stebbings55:47

        What percent churn at that moment?

    16. JGJean-Denis "JD" Grèze55:51

        30% right off the bat. Yeah. Bang on.

    17. HSHarry Stebbings55:53

        How do you get that down?

    18. JGJean-Denis "JD" Grèze55:54

        You gotta be willing to take a hit.

    19. HSHarry Stebbings55:55

        What's the biggest internal product disagreement you guys have today?

    20. JGJean-Denis "JD" Grèze55:59

        We have product market fit for, like, some purely personal use cases that we didn't expect, you know? And the problem is it's like-

    21. HSHarry Stebbings56:07

        Like, like what?

    22. JGJean-Denis "JD" Grèze56:07

        It's like product market ... Like families, like parents. Parents and families, there's, like, tremendous product market fit for Town there. Because schools ... You don't have kids, right? You have, yeah, you have a girlfriend. You don't have kids. So, um, schools, in America at least, send a lot of emails, and they have a lot of, like, portals where things have to happen for sports leagues and they're like this kid's reports, and there's a lot of scheduling for kids that has to happen for, like, haircuts and summer camps and all of these things. And our product's, because it's really good at email and it's really good at the scheduling stuff, it, it really has tremendous product market fit for families. Um, and we have, like, a marketing ... So it has the product market fit. The question is, like, do we market to this group, and do we spend time on it? Because it's a great group, and it has willingness to pay, but it does not have the willingness to pay of a mid-market firm, right? It's just a, it's a different look.

    23. HSHarry Stebbings57:03

        It also doesn't have the, like, virality and expansion of moving across an org. And, like, yes, you can go across parent groups, but it's not like you get into a Revolut, and then Revolut c- has 7,000 people. That's a big fucking-

    24. JGJean-Denis "JD" Grèze57:18

        Yes

    25. HSHarry Stebbings57:18

        ... expansion if you can nail it. Yeah, all the people in Sonoma who have kids under five, great.

    26. JGJean-Denis "JD" Grèze57:25

        Yes, but it jumps ... I agree with you. And so I think the, the discussions that are being had is like it's good growth, but it doesn't ... It's, it's not for monetization. You have to believe something like those people will then bring their Townie to work use cases in places that we wouldn't be accessing as quickly, or you have to believe that the word of mouth is incredibly powerful. And so the f- the interesting thing about the parents for us is they do talk about the product a lot in, like, like their WhatsApp parent groups, in, like, Facebook groups and places like that, right?

    27. HSHarry Stebbings57:58

        There, there is network virality that I complete-

    28. JGJean-Denis "JD" Grèze57:59

        Yeah

    29. HSHarry Stebbings58:00

        ... especially on the community groups. Yeah, yeah.

    30. JGJean-Denis "JD" Grèze58:01

        But, but I think, you know, we can't do all the things. So, you know, to your point, it's like, well, we're a monetized platform, right? And so our metrics are, AR growing month over month, right? Revenue is what matters to the business. And so, you know, it's why it's an argument 'cause you're ... When you find product market fit somewhere you don't expect, you have a few choices in life. Like, one choice is you're like, "I love these users, and I love parents. I, like, love the users," right? I love the use cases. The use case isn't the value super clear, but it doesn't ... It's not fully aligned with how we've thought about the business growth. But, but I think that's why we're having inter- interesting thinking internally 'cause we're like, "Is this ... If we look around enough corners, is this worth it? Or, you know, should we, should, should we be more focused on our existing strategy?" Which is fine. When you build a company, you learn things from users, and you gotta make the right decisions.
19. 1:00:01 – 1:03:41

    Is the AI Assistant Hype Justified?

    1. JGJean-Denis "JD" Grèze1:00:01

       I mean, for us, we have the revenue and the growth. I don't know what, you know, competitors' growth is. I think it's ... If there is a path, if you can believe that some of these companies can get to tens of millions of people in the products in an area where the product will be The entryway for people to do digital things, right, that they're doing in apps on their phone right now. If there is a winner there that comes out of the startup universe, there is like a giant company to be built, right?

    2. HSHarry Stebbings1:00:31

       The billion-dollar price for the new round, did it start there or did it get ratcheted up and up and up?

    3. JGJean-Denis "JD" Grèze1:00:38

       I, yeah, I, I can't c- I can't deny or I can't confirm or deny.

    4. HSHarry Stebbings1:00:43

       Oh, right. [laughs]

    5. JGJean-Denis "JD" Grèze1:00:45

       I love, I mean-

    6. HSHarry Stebbings1:00:46

       I'm being [laughs]

    7. JGJean-Denis "JD" Grèze1:00:47

       Harry, I will not let you take-

    8. HSHarry Stebbings1:00:48

       D-

    9. JGJean-Denis "JD" Grèze1:00:48

       ... any in, any moments from me away from me.

    10. HSHarry Stebbings1:00:51

        Uh, you're not commenting anything, and I'm not even prying, but like the, "Oh, Inda is doing it. Oh, Inda is not." What's great for you is it's all just PR. Like, your name's just everywhere. It's pretty good, man.

    11. JGJean-Denis "JD" Grèze1:01:03

        You're, you're amazing at trying. The, uh, I'm old, you know? I'm, I'm like, I'm like 47 years old. I don't even know how old I am. That's how old I am. When you know you're old when you don't remember if you're turning a certain age. So I'm 47 turning 48 in a few months. And, um, and I've been around for a while. And there are times in my life where publicly I've done things that I'm proud of, and publicly I've done things that aren't, that I'm not proud of. And, um, so the reason I mention this is like I, I don't think it aligns with my value system on anything to like create PR just for the business. I think I want the PR to be created by my users because they love the product. So if you ever see any news about Town that, you know, good or bad, that is ... It's ... I'm not out there creating the PR. That's just not, and it, that's not just my way of operating.

    12. HSHarry Stebbings1:01:51

        I, I, I think that's, I think that's a mistake that respectfully-

    13. JGJean-Denis "JD" Grèze1:01:54

        I know

    14. HSHarry Stebbings1:01:55

        ... I would push you to change.

    15. JGJean-Denis "JD" Grèze1:01:57

        Probably.

    16. HSHarry Stebbings1:01:57

        And I, yeah, and I would say like l- look at a Whisper flow as an alternative. Not a hugely dissimilar, uh, PLG motion. Um, in all candor, I think they've done a brilliant job at generating PR themselves through their own content, through content that-

    17. JGJean-Denis "JD" Grèze1:02:11

        Wow

    18. HSHarry Stebbings1:02:11

        ... their users produce. Sure. But, like, content is a hack to customer testimonials.

    19. JGJean-Denis "JD" Grèze1:02:17

        Totally. I, I, I agree. I just don't think a fundraising story, you know, is part of the universe of things that I would wanna like create a PR moment out of. Um, so like not the kind that you're referring to from, from, you know. So I w- I do wanna make a point on the, on the fundraise, 'cause I, this is as an angel. It's not about, right, it's not, not about us. The fundraising in general is ... You, you kind of mentioned, you know, if where it started. Like there are, there are these companies out there, this goes back to the ethics and who I, who I am, like I have seen deals where it's like I invested like 200 and then the announcement is at 500. And then what you learn is that like, you know, they raised 65 million and like 5 million's at 500, and the other 60 is at like 200 or 300, right? There's a whole lot of that happening for sure in the Valley. Personally, I, I don't think it's ethical. I don't think it's ethical towards employees, you know, most of all. If you're not ... I mean, maybe when you hire someone you tell them for sure. Because it's like the dilution wasn't that number one. It's not where most of the demand was, right? It's not how you should be pricing people's offers. You can't, I don't think you can look at someone in the eyes and say, "An investor that made 80% of their investment at like a 200 or a $250 million valuation, but hey, I, I'm, you know ... They, they, they put the last 20% at 500, and that's what I'm gonna say." I just feel, like I just don't like that. I don't feel

20. 1:03:41 – 1:15:49

    Quick-Fire Round

    1. JGJean-Denis "JD" Grèze1:03:41

       like it is right.

    2. HSHarry Stebbings1:03:41

       Uh, dude, we're gonna do a quick fire round, okay? What is your best angel investment?

    3. JGJean-Denis "JD" Grèze1:03:48

       Oh, it's either Base10 or Modal right now. Those are the first two that come to mind.

    4. HSHarry Stebbings1:03:53

       What is the bull case for Town being a $100 billion company? What is needed to happen in that world?

    5. JGJean-Denis "JD" Grèze1:04:00

       I think if we can get about 10 million people paying for the product, we, we can get to that.

    6. HSHarry Stebbings1:04:07

       Is that it?

    7. JGJean-Denis "JD" Grèze1:04:08

       I mean, our ... Yeah, we make over 700, $700 per year per user today.

    8. HSHarry Stebbings1:04:14

       How does that compare to Dropbox? 'Cause Dropbox must have way more than 10 million users.

    9. JGJean-Denis "JD" Grèze1:04:19

       You have to have the growth, right? I- I- it couldn't be ... I couldn't terminal a 10, right? I'd have to believe that I can keep getting a, a good rate of growth. The problem ... I mean, I, I'm pretty far from Dropbox. Been a long time since I worked there. But I think there was a huge free, huge pack of free users that were very costly on the cost side. And then on the paying side, I don't remember if the number was like 10, 20, 30 million, but it did, you know, it did flatten out at some point, and there was no way to generate more revenue or growth from the users. I think what's different on the AI space is, like I think you should be able to, as you do more ... Like once you have a company on a platform using your platform as the core part of where AI works happen, you can, you, you should be able to generate increasing revenue as, as the token spend goes up.

    10. HSHarry Stebbings1:05:00

        Who would you most like to add to your board who you do not have?

    11. JGJean-Denis "JD" Grèze1:05:04

        Ooh. I think for the next board member I, I would love someone that's kind of CFO-like. Late stage our business is gonna be like the economics are gonna have to be really, really good. I know this sounds weird, but I think you need, if you have a board member with real operational experience on the, on the f- on the finance side, it's gonna be very helpful as you scale this kinda company. There's just a, there's a bunch of stuff, like we're gonna have to buy compute at scale. We're gonna have to be like very, very good about thinking about token spend. So it's, it would be someone with that background. I know you're making faces. You're like, "Hmm." I just, I'm not a big-

    12. HSHarry Stebbings1:05:36

        That was s- that was sooner, that was sooner than I thought though. I, I get that need, but I thought that would come in a couple time.

    13. JGJean-Denis "JD" Grèze1:05:43

        Maybe. But I think, I think we're in growth investor land for the, for the next round. So I think once we're in growth investor land, they, they will, they will ask for me to have ... They will want for fundraise metrics that, you know, I can really defend. And I think having someone with that background will be helpful.

    14. HSHarry Stebbings1:05:58

        Why don't you subsidize completely? I'm, I'm being serious. W- you could raise another 200 million more, and growth is everything.

    15. JGJean-Denis "JD" Grèze1:06:04

        Yeah.

    16. HSHarry Stebbings1:06:04

        Why don't you just go, "Fuck it, burn the boats"?

    17. JGJean-Denis "JD" Grèze1:06:07

        It's a g- it's a good question. And I, I think I, I would be lying if I said there aren't mornings where I wake up and I think about it. I believe that to prove value on the business side, you must make your customers pay. So I do think there might be a world where our, where the PLG part is much more subsidized, but as soon as you get three to five team members, I really wanna make money on that side. I really wanna make sure I'm delivering value. I'll give you a, a story. We, we... When we launched the product initially, right, for the first five months, not launched, but we were, like, in private beta, and then we opened the beta, we didn't have pricing. Okay? There were users who were spending, I shit you not, like $2,000 of compute a month, $4,000 of compute a month. There's someone on the platform who'd spent, in five months, was, like, $26,000, okay? And because there's no pushback on the token spend, right? There's no pushback at all. So I know, I know, you're making face. It was one... That we would, like, call them, and we'd be like, "Look, like, you know, let's, let's figure it out. Like, you know, you're using the platform..." Because you could just create, like, routines to automate more and more stuff, but is it really bringing value to them? So anyways, the reason I, I mention that is I'm a big believer that getting pushback from the market about where you're delivering value and where you're not is really, really important. There would be ways to subsidize and do that. So for example, I could make the plans much cheaper. I could make them free. I could give free tokens to businesses. But what I've learned is, like, on the business side, they also don't like it if you don't charge them because they don't know how much it's gonna cost one day. They wanna know how much it's gonna cost one day. You can't sell to a 500-person company and be like, "Yeah, just use my product for free internally," until you get a bunch of usage, but then one day I'm like, "I'm gonna turn it off," and all your business processes are running on it. So we, the way we've approached it is on the growth side, we may or may not subsidize more, and 'cause we're investizing growth. But I really want a real business when a company is on this product, and we have a real business when a company is on the product, and I'm very proud of that because I think that is the ultimate test of whether you're building something successful if you're not a pure consumer company. And I'm doubtful of pure consumer ad backed for AI for a couple reasons. Like, I think the, the tokens are way too expensive to do ad back now. And then number two, there's an incentive problem with ads, and I think people are gonna want assistance that, that are theirs, that are not being polluted by out- outside incentives, like ads, into the trajectories that they give you, right? So if you ask to book, like, a flight and, you know, it uses an airline that is, like, paying for that flight to be recommended to you, that doesn't feel good, right? So I'm a big believer that actually the economy around assistance will be paid for, and so I just wanna pay for it as soon as possible.

    18. HSHarry Stebbings1:08:52

        What person, if when you open Twitter, would you be most thrilled to see love Town?

    19. JGJean-Denis "JD" Grèze1:08:58

        [chuckles] Elon Musk.

    20. HSHarry Stebbings1:09:00

        [laughs]

    21. JGJean-Denis "JD" Grèze1:09:01

        'Cause he has a competing product, and it's Elon Musk.

    22. HSHarry Stebbings1:09:05

        How has your hiring process changed in an AI world?

    23. JGJean-Denis "JD" Grèze1:09:08

        Well, you know, we have a weird hiring process. You wanna know a fun thing about our hiring process?

    24. HSHarry Stebbings1:09:12

        Yeah.

    25. JGJean-Denis "JD" Grèze1:09:12

        If someone on the team has worked very closely with someone else, we don't interview them. Yeah.

    26. HSHarry Stebbings1:09:20

        But why?

    27. JGJean-Denis "JD" Grèze1:09:21

        If it's a top person. Because why would I? I just sell. I will just sell. So it, it's, it's, it's very rare that it happens, but it has to be someone that they've worked extremely closely, like, literally, like, next to, and they're like, "This is one of the best people that I've ever worked with." And for, for, like, eng, we're just like, "Let's go." Because it's, it's... I know it sounds odd, and people are gonna, like, comment, like, "This guy's a total idiot," but a person that I trust that's great on my team, like great on my team, they tell me this other person is, like, one of the best people that I've ever worked with, and then I'm gonna make that person, like, spend eight hours doing stupid whiteboard interview or, like, pair... Makes no sense. So either I don't trust my employee. So there's culture match. So we will be like, "Hey, come in, spend some time with us. You know, like, you can code with us if you want to." We have to sell 'cause we look... You know, if you don't interview somebody, they're also like, "What kind of clowns are you? You're not interviewing anyone?" So we will, like, allow them to get signal about us, but we are not evaluating whether they can do the core role.

    28. HSHarry Stebbings1:10:17

        Brian Singerman, who invests in funds and then invests in the companies beneath those funds, has a rule that if the manager's like, "Balls to the wall, I am all in on this company," he'll automatically write the check. Kind of the same. You trust the person, you trust the layer beneath them. So I totally get you there. Um, what percent of developer salary do you spend on tooling? So Marc Benioff said, "At Salesforce, we spend $300 million on Anthropic." They spend $6 billion a year on eng, 5%.

    29. JGJean-Denis "JD" Grèze1:10:44

        I mean, the run rate's at least 75K per employee, per engineer.

    30. HSHarry Stebbings1:10:49

        Split between Core Code and Cursor?

Episode duration: 1:16:00