Founders in Arms
00:37:40 · Immad Akhund, Jean-Denis Greze, Rajat Suri
[00:37:40] S0: And we'll [00:37:41] S2: do, like, market map for you. And then you can be like, hey. I want you to put in this spreadsheet on Sunday a list of 200 new candidates that match the job description, and I'll review it manually. And then after you're done reviewing manually, you [00:37:54] S1: can be like, hey. [00:37:55] S2: Can you start emailing the candidates to see if they wanna talk to me? And... [00:37:59] S1: Helps solve business problems. Yeah. You know? [00:38:01] S2: And it... It's like it does all these kinds of human workflows that are coming in and out of email, your SaaS, your system of record, your calendar. Like, the goal is to automate as much of that as possible. And it's pretty simple to use, so non technical users are able to create those kinds of automations without a ton of work. [00:38:17] S0: I do feel like this is a tricky thing for you that everyone has, like, AI assistant and email, and you have to, like, try to differentiate, like, you know, where you are amazing when, like, everyone seems like they're building horizontal tooling. [00:38:33] S2: Yeah. The the thing is if you talk... If you're, like, in Cleveland talking to people, they're, like, [00:38:39] S1: so far from this. Yeah. It's an s f... Yeah. Everyone's gonna work [00:38:42] S0: on Everyone. This is weird. Yeah. No.
00:38:44 · Jean-Denis Greze
[00:38:44] S2: No. And it's it's totally okay. But, like, the the the the the question that I ask myself is the capabilities problem. And so I actually think on the capability side, meaning what can the assistant do, all the assistants will be able to do the same things. Because the harnesses will normalize more and more. Third parties will provide you the ability to, like, you know, shop online, order flights, deal with spreadsheets, all of those things. So right now, we're still in a period where the hardest matters, but long term, it doesn't matter that much. I think the places that are different is, like, what users are you picking? How proactive are you gonna be for those users because you're willing to go deep in the use cases? And then how do you market? Like, what what is... What is your go to market strategy? And so the reason why Town works is we focus on personas that are basically not Silicon Valley personas just by having a lot of users here. And our entire onboarding experience and our marketing is very good for those personas. So I I would say the recruiter example, it's like, if you're a recruiter in the Midwest as a small firm, and you see our ads or you see a LinkedIn post from someone who uses a product today, you click on it and you sign up for a product, the onboarding experience will make you feel much more like, oh, this is AI for recruiters. And you've actually never had AI for recruiters. And for them, it it doesn't feel like an AI system. It feels like an AI system for recruiters. And what we're working really hard to do is to make people feel like that in as many roles as possible. And the mix of product and marketing around that is why it's working right now. Will it work long term? And we'll see. Right? Like, I'm in the arena, and that's the best I can do.
00:40:28 · Rajat Suri, Jean-Denis Greze, S4, Immad Akhund
[00:40:28] S1: Okay. Alright. [00:40:31] S2: It's a [00:40:31] S4: question for all three of you at some level. The cost of development is close to zero right now, as you're saying with codecs, cloud code, and stuff. But when you're trying to build a product, it's about the product taste, it's about user feedback. What is the equivalent of plot code or codecs for that? How can you accelerate? And I know we have second brain. We are trying a few things. But is there anything that shortens the cycle in terms of product building or... [00:40:55] S1: Do you [00:40:56] S0: wanna repeat the question? I think [00:40:57] S1: the question... Let me see if I understand it first. You're saying the... Solving the problem of what to build, can AI help with that? Is that the question? Yeah. And [00:41:07] S4: how do you make sure that you're not adding another key in the... Not the keyboard, but instead [00:41:11] S2: coming up with an iPhone? [00:41:13] S1: Right. But how do you know if you're building the right product?
00:41:13 · Rajat Suri, Immad Akhund, Jean-Denis Greze
[00:41:13] S1: Right. But how do you know if you're building the right product? [00:41:19] S0: I mean, there's a reason humans still have jobs. I do I do think that, like, if you think about the product development life cycle from, like, come up with an idea, you know, or maybe even before that, like, talk to talk to customers, figure out what's the idea, figure out, you know, how are you gonna build it, what's in scope. Yeah. The actual, like, engineering part was the expensive part previously. Right? Like, you've spent maybe 80% of your time on the actual, like, coding, and now that's maybe, you know, 20% of the time. So, yeah, I think the rest gets more important. There are things that help with AI. Well, AI helps with the rest of it. You know, you can look at all your support tickets and, like, summarize it better. Maybe you can plug that screen into kind of linear, and then, yeah, I can have a go at, like, trying to do a prototype or something like that. So there are, like, things that are speeding up our process, but it's definitely not uniform. And, yeah, what we think about it, I guess, everyone's thinking about it. So, you know, what is the next biggest thing that you can credit with these 90% of the new coupon? Kinda squeeze all the bottlenecks. But it's not... I don't... Like, there's a magic answer to it necessarily. I think one thing that I think is important is that engineers do more and more of those other tasks. So, like, previously, because coding did 80% of [00:42:42] S2: the time, you would want
00:42:42 · Jean-Denis Greze, Immad Akhund
[00:42:42] S2: the time, you would want [00:42:44] S0: an engineer to be kinda distracted by some of those other things. And now, like, you know, that's the small percentage you want
00:42:51 · Jean-Denis Greze
[00:42:51] S2: the engineer to be kind of full stack across the in the life cycle. Yeah. I think... By the way, another answer for... That we are seeing more and more is that it's difficult for users to discover features in these products because the products can do more and more and more. And whether the features are just like, hey, if you write the magic words in the box, it'll do this thing that you didn't think it could do before, Or the feature is, like, in a menu somewhere. And so, like, the the odd the odd thing for me is, like, hey, it might actually... You you may not want unlimited features. And even being better maybe listening and understanding the full range of features is not is not that helpful because humans that are clicking on things are still kind of more limited in what they can click on. So, like, my mental model is if it's a headless feature and I can use existing incantations and make sure they happen, then you kinda almost don't wanna ask the question if it's the right feature because it's headless. You wanna get it in your product, and you want to let the AI decide whether it's useful. And and if you think that way, and your heart is as good as discovering that there's a problem and this new feature solves the problem, it doesn't matter how sloppy or bad the feature is in a way because you're gonna recognize from product analytics that there's demand for it, and then you can iterate and make it better and better. And then for the rest, there are things that humans interact with directly. Then I think taste matters almost more because everyone's like slopifying left and right. You know? It like naturally slops because you can build more things. And so I think if you're very tasteful and thoughtful about what is your strategy that gives you competitive advantage, like what is what is the set of things that if you had, your your users would be super happy, and then you build those in a delightful way. It's all taste. Right? Because it's a human that's judging whether it's useful. There, I think it's not that different than it it was before.