Why 90% of Founders Build Startups Wrong

Yeah. I think it’s it’s just this concept we were talking briefly earlier around, you know, Math Olympiads is just a it’s a type of math contest, and there’s nothing really fancy about it other than it kind of collects a lot of really smart, talented younger people together. And what I’ve noticed is that a lot of the folks in my generation have, you know, gone on to start companies and done very well. And I’m now in touch with a lot of the younger generation as well, who are just really smart and really ambitious.

And it just turns out that these folks do very well in startups. And it’s maybe not a surprise on the surface level. But historically, a lot of, you know, the people that have performed the best at, you know, math contests and things like that have generally gone on to do research and quant trading and things like that. But I think we’re seeing that if you can really unlock that level of, you know, reasoning capability, and just like smartness, and combine it with folks that can, you know, teach them how to sell or how to build a company, I think that’s a pretty good combination.

So that’s something I think is a bit underrated right now. I’m really excited to spend as much time as I can with the younger generation and see if I can, you know, push them into doing more startups, doing more sales. I think that’s that’ll be really, really fun.

Yeah. Let’s do it. I think that would that would perform quite well. I mean, at least at that scale. Think I once you scale it, it’ll it might become a little bit harder, but for sure.

Lowkey was the first company I did out of school, so I’d not known how to do anything beyond that. I mean, when I was in college, I interned at, you know, places, quant trading and startups and stuff like that. It was my first real job. It was a pretty tough journey, honestly. When you’re first starting, you don’t really know that many things. And so we we went down all these wrong paths, and it was a bit of a grind. But what

By far, it is just over intellectualizing things, I think it’s just a very common pattern where you start and you’re just like super excited, you’re listening to podcasts like yours, or just reading all these articles, and you just it’s like really easy to form these narratives in your mind, when in reality, this doesn’t really matter that much. And there’s just way too much noise that way. And so I am happy to talk about like, what I think the right way to build companies is now. But the issue was we just over intellectualize, they were like, this must be a good idea because of x, y, z, like, there’s these trends in the market and so on.

And we just built things that people didn’t really care about. So there were a bunch of cycles of that. And those are very demoralizing, because you spend a ton of work and a ton of effort, and you’re like pulling on nighters. And then at the end of it, just like literally nothing happens. That was the issue. But I mean, fortunately, towards the end, we, I mean, both from just like persistence and also just lucky, we launched stuff that actually took off. And we’re in the right place, right time as well.

We got a really good acquisition offer and that’s what happened. And then to answer your question directly, I think the second time around, you’re a little bit more grounded and you’re also swinging a bit larger because you feel like you have a win under your belt. And so if you have another one of those it doesn’t really matter. And so you’re you’re just gunning for for bigger ideas.

I think what I’ve noticed is that generally when you try new things, where you learn new things, the rate at which you’re able to discern good versus bad improves a lot faster than your actual skill level. A classic example of this is like if you’re learning piano or something, right? Like the rate at which you can tell what is good music versus bad playing is it can actually grow pretty quickly, but your own abilities will be a lot slower. And so what that leads to is generally people quitting early or just not seeing it through or just getting demoralized.

I think there’s a bit of that with startups as well, where because you’re able to read all the articles about, know, great founders that have, you know, great stories, or you’re listening to podcasts, you feel like you can discern good from bad really quickly. And then when you’re doing the work, obviously, it’s a lot messier. And so you’re in this situation where it’s really easy to not pursue things, but it’s really easy to also just like follow hype y trends and stuff like that. I think the number one thing that people should do is just ignore all of that.

It’s off obviously useful signal, but it’s just way too dangerous to read too much into that. I think the right thing to do is you just have to dive into it and you learn through just reps. And for us this time around, the reason why we’re able to get off the ground a lot faster compared to my previous company, would say is just that we did that. I think people were telling us at the time like, oh, there’s all these AI ideas, they’re going be way too crowded, don’t do them.

Or like, maybe there’s this idea over here that could be useful. And we just decided not to listen to that. We just spent a ton of time talking to customers directly and asking them, alright, what’s useful to you? How much would you pay for this and so on? And we just got a lot better at doing that discovery. I think that’s what you have to do.

In the early days, execution should help you find the right markets. Because if you’re actually executing your discovery well, you should be able to discover, okay, which markets actually will be real versus not. I mean, that’s the exercise we went through. Right? We were obviously, literally everyone else at the time, really excited about LMs and AI agents. And when we went in and talked with all these potential customers, we’re talking about all sorts of their use cases. And at the end of it, a lot of these discussions ends in, okay, well, this is great.

Like, it sounds like you’re really interested in this. How much should we pay for it? Like, what does this mean to you? And you get into this, I don’t know, like budget tights, like, maybe, like, a $100 a month, like that sort of thing. And then you’re like, okay, yeah. So through discovery, we realized that those ideas are probably not good. And when we when we started talking about, you know, the CX space and actually deploying conversational AI agents to actually talk with customers, people were able to justify that way more, right?

They’re like, oh, wow, yeah, this would be useful because I have 400 support agents right now. And you guys can do something good. Be willing to pay you, you know, 6 figures the bat. And when you’re at $0 in revenue, you’re like, oh, great. 6 figures. That’s that’s awesome. That’s the general gist of when I talk about execution. It’s like it’s really going through that hard work.

I think it depends on the market. I have a huge respect for the Harvey team. They are almost exactly a year older than us. And so they’ve been very gracious about, you know, teaching some of their learnings. And I think it just depends on the market. So I think their market is one where it’s a lot more self contained in the sense that like, if you get the big players, there’s only so many law firms out there. Like people will see that and like, oh, that’s great.

And like, kind of you join in. And so in that sense, I think speed matters quite a bit in kind of selling the vision and so on. Our market, for example, is just a lot broader. And so you have to balance it a little bit. You can’t just go in and sell the vision because no one needs to buy into the vision. I think everyone already believes the vision. And so then it’s more about showing the results quickly and being able to demonstrate that, hey, there is ROI here and that, you know, we’re not selling vaporware.

You can actually deploy it and and see results.

In our space, there are actually many players. So Sierra, we have a lot of respect for them. I think they they’re a strong team. Salesforce offers different advantages. I would expect a Decagon and a Sierra to be able to execute faster on a product side than Salesforce, but Salesforce has, like, so much distribution. So we’re always thinking about, like, the different generations and, like, how to counter position. Yeah. I would say in in some sense, there’s a land grab because there’s just, like, so much excitement in the space, and so you wanna get as many customers as possible.

Yeah. Zero to one is actually just me and Ashwin. So it’s just the two of us. And the advantage there is that we’re able to move much faster. So we’re doing all the building and all the sales and all the talking and like, it’s just like two people doing it. So you’re able to move a lot faster. The biggest learning I mentioned before was that we really just went deep on discovery. When we’re talking with these customers, it’s really easy for people to express their excitement for something because they’re on the call with you, they want to give you something.

But oftentimes the excitement doesn’t necessarily translate into, you know, revenue or actually like some sort of commercial agreement. And so I think we were pretty good at executing there. And then from there, it’s just like, can you build fast? And when you’re in these comparisons, and they’re like, oh, okay, there’s this, you know, two person company that’s not really known, but I’m going to compare that to some of the older solutions, can you outperform that? And so that’s, that’s what was important in the zero to one phase.

I would say even at one, like, I wouldn’t say we were that well known. I think we had started to become a bit more better known in the tech world. If you compare it to now, I think now we’re like way more well known. And so it’s just kind of the spectrum where you have to balance your strengths at any one time.

The seed round we raised before we had any ideas. So we raised from Andreessen Horowitz. They’ve been awesome partners. They backed my last company as well.

I think it’s more important than people give it credit for. A bunch of reasons. I think one is that for hires, it actually matters a lot. And you think it wouldn’t, but if you just pull the average, like, you’re a really good engineer out there, like, they do care about who your investors are.

I mean, I just think that most people care. If you’re responsible about your career, like, you’re gonna look at all the different factors. And maybe the sort of kernel of truth behind that quote is that that shouldn’t be the only thing they care about. But if you’re a reasonable, like, person and you think a lot about your career, you probably do care about the investors that that are supporting the company.

The other piece is just with customers. Right? Like customers do care, especially in the early days. It gives them a little bit of validation that, hey, these people are not just gonna be frauds and completely just waste my time. That’s helpful. I think the thing that is overrated, and I don’t know if this is a hot take, but early on the brand name VCs will sell their like platform and say, hey, we have like a great platform as well that’ll help accelerate. I think it’s basically impossible for any VC to help accelerate the process of getting to PMF.

I think the issue is that VCs will sell that they can, but that’s not possible. On the flip side though, I would say we’ve really found the platform teams that are investors quite useful, but that was like after we started accelerating.

Yeah. I I think we were just in a very fortunate situation. Like, all of our rounds were like so oversubscribed that we just didn’t really care. I actually think given a choice, you should have different investors for the C and the A because you just get two firms instead of one. They bring different things. So we we wanna excel for the A and, yeah, it was a it was a good choice.

I think it doesn’t really get easier. Your the stuff you’re solving for just changes a bit. I mean, product market fit, don’t think there was, like, a one day where we’re like, oh, we didn’t have product market fit yesterday, and now we do. It was it was kind of a gradual process, and the work becomes less of the discovery building type work and more of the operationalizing and scaling work.

I would say for AI, mostly it’s starting fresh. And the reason is that having the older way of doing things actually serves as baggage because you have so many customers already locked into that approach. And because you have a lot of customers, anything you build has to be compatible with all of them. And so it just always inherently gives startups an advantage. And that’s why when you have these like big technology shifts, it just opens up a big area of opportunity. Because it’s not like the old players don’t know that AI is here.

Of course, they are investing in AI. Of course, they want to be AI friendly as well. But it’s just tough for them to compete head to head against more agile things that are AI native.

Sure. Yeah. I’ll give a very tangible example in our space. So if you just think about automation in, you know, customer experiences in the past, there’s been a long history of it, right? You have these phone trees that no one likes, and then there’s chatbots as well. And generally the way you build them is the classic SaaS approach, which is you give people like a framework and sometimes it’s like a configuration language, like what Salesforce does, where you just have technical people just building stuff out.

There’s a couple of issues with that. One is that as things get really, really complex, the rate at which you can iterate becomes slower. And the other thing is that it often gets bottlenecked through engineers. So you need technical people doing stuff. The main thing that AI unlocks is that you’re able to democratize that a lot and empower a lot of the non technical business users to use natural language as the medium to build things and iterate and so on. So what we’re seeing in this new generation is that, one of the biggest innovations that we brought that people really liked was we have this concept that we call AOPs, Agent Operating Procedures, that are mostly natural language, and that’s how you build AI systems.

And if you compare that with the previous solutions, right? Like none of the previous work you would do to kind of create all these configurations and stuff, none of that really translates. Like that doesn’t really help with this new approach. So that’s what I mean by kind of levels the playing field a little bit, because if someone were to really just sit back and be like, ignoring what I built so far, what is the best way to actually bring AI to these teams? The answer probably is like, okay, have to throw away most, if not all of what’s been built so far.

So that’s that’s why the the playing field is leveled a little bit

I would say that’s already happening in our space. So the reason why the folks in our space are growing fast is just maybe two people, or that you’re able to close larger deals. And that’s not to say you’re able to capture the entire human labor spend, but the sort of benchmark is pegged a lot higher. This is an interesting point because if you just look at software purely, oftentimes you’re just compared to what the existing software solution was, or maybe you’re compared to what your costs are.

And so if you’re building, and this is not to bash the infra solutions, but oftentimes when you’re building stuff that’s kind like in the middle of the stack, you’re judged on, okay, well, the models cost this much. And so we’re willing to pay you that plus like some percentage. But in this case, when you’re in the application layer, if you’re building enough of a product, you’re more benchmarked against what is the business problem you’re solving. And the business problem you’re solving is going to be way bigger, right?

To your point, it’s, you’re saving human labor, you’re, you’re transforming the way that their customer experience works. And if it’s good, you should be able to increase the revenue. So that is really helpful here. And that’s, I would say that’s probably why this market has been so good so far is that people have been able to make that justification. It’s, in my opinion, one of the few markets right now that has true PMF with AI agents.

I think it just goes into sort of enterprise top down sale versus more of a PLG bottoms up motion. And there’s trade offs. Right? Like, I mean, those are amazing businesses. If you’re a PLG bottoms up motion, you’re just growing way faster. You’re just it’s just like, boom. Like, I don’t know what the revenue

Yeah. Exactly. But but the flip side is, like you said, right, because it’s a product like growth, there’s a lot more downward pressure because it’s much easier for competitors to come in and so on. And that’s not to say there’s no downward pressure for enterprise sales. Of course, there is as well. But it’s a lot easier to kind of go in and you’re just doing a lot more than just giving them a product. You’re helping them with the implementation. You’re really crafting the agent to be super customized to them.

And that’s what unlocks some of the higher value.

It’s tough to say just as a rule of thumb, but if I had to give a rule of thumb, human labor is generally like an order of magnitude larger than software spend, like 10 x or more. Of course, you don’t have capture all of that, but when you’re kind bringing in software, if you can give them like a three x to five x ROI, that’s very compelling. And so what that translates to is like, you know, roughly three x multiple on the value you’re able to capture there.

And so if you if you look at a lot of our customers, for example, they’re spending more on the AI agent than the previous CRM software for support. And that’s because the the value is just a lot higher. You’re able to just, like, solve the issue completely instead of, you know, just giving software for humans to use.

Not really, actually. So I would say the benefit in our space is that C suites already want to adopt AI. Like, no one needs to be convinced that like, hey, we need to use AI. And oftentimes when they look internally in the org, okay, where can we adopt AI? The obvious ones tend to be, you know, customer service and then maybe like coding, like you said, right? So those are the two obvious ones. That is helpful for us and that we don’t have to go in and like try to pitch that, Hey, you should be investing this, like people know.

And so then it’s more, it’s on us to really demonstrate the business case, right? It’s like, Hey, we have shown that we’re able to resolve this many conversations and we’ve shown the rate at which you can improve it and add new workflows and iterate on it is really fast. And so in our first year, we think that this will happen and we’ve proven it in this pilot and the customer satisfaction has also gone up. So it should be a no brainer. That makes it easier.

That’s very different for a company to company. Just as you imagine, like the distribution of their customers and so on. Right? But I would say that like a good bar when you’re fully up and running is like sixty, seventy, 80 in that range. And it just depends on like what sort of things the AI has access to. Like, if it’s able to take a lot of actions and resolve a lot of things, then of course, that’s that’s gonna be higher.

I think there’s a couple of nuances to that question. I think this space, the reason why it’s exciting is not just that that number goes up. Like, of course, that number should go up. I think in three to five years, it’ll probably creep towards, like, just consistently 80 to 90. Right? Because you’re able to just so easily allow the end users to capture more logic and teach you new things. And I mean, that’s the sole purpose our company exists. But at the same time, the sort of scope of AI agent will expand.

I think that’s the exciting part. And that’s why a lot of senior leaders are excited is that when you talk about 70%, right? It’s 70% of all the support inquiries. Like, hey, I need to reset my password or I lost my item. Like, I need a new one. Or I have questions about my loyalty points or whatever. Right? Solving those is obviously great. But if you had a really, really good AI agent, it becomes almost like a new conversational UI of your brand. It’s like a concierge almost where you just talk to a lot more about all sorts of different things.

Like maybe you start talking to it before you become a customer, or it’s helping you kind of learn about the product and that leads to more revenue because you’re opting into a new product line, or it starts to be more proactive instead of reactive. And so that’s where the excitement is. And so I would say in three to five years, that’s going to become a lot more real where a lot of orgs will have implemented things like Decagon and resolve a ton of conversations. And then the next step is like, okay, how do we be more forward looking and how can we be more revenue generating?

At that point, the number still matters, but people will look at other stats as well. Like, what is my retention rate? What’s my conversion rate?

Model performance is obviously quite important. I think maybe the difference between our business and something like Winserve is that the reasoning capabilities matter a little bit less. I would say the models as they stand today are generally good enough at solving most of the meaty inquiries in customer service. It’s like instead of reasoning, it’s more about instruction following. Can I follow all the rules correctly? Can I execute this workflow correctly? Can I take this AOP from Decagon and really make sure that there’s no steps missed and so on?

I would say the models are generally good enough at that. So then this challenge becomes, can you orchestrate it well? Can you figure out, like, which models are the best at which things? Can you make sure that latency is optimized? Can you make sure that consistency is optimized? I would probably say less less of an issue in our space.

The way I would describe it is it’s all about finding the right level of abstraction. So if you’re being super customized, right, you would basically go to a customer, you’re like, all right, what is the conversation or the types of conversations you have? And you would just like code it from scratch and be like, all right, here’s an agent for you, right? That’s not going to be a huge business because it only matters to that company. So if you zoom out one layer and you’re like, okay, maybe now we’re solving all customer support inquiries, but for your industry.

And then you might be able to build some level of abstractions so that, you know, they can go into your product and fill around with some things and build their agent. And now they’re able to do all the common workflows like refunds or chargebacks or whatever. And then we zoom out one more and that’s where we’re at is, okay, now we have this abstraction layer AOPs that allows you to build essentially any sort of conversational support inquiry. And when I say support, it’s the dynamic of the users coming inbound to you, right?

And you have a procedure that you’re executing. Now, I think we have a pretty good abstraction layer for that of how you can just use that framework and customize it to any sort of use case that is, you know, support specific. And so to make the jump to what you’re saying, you have to go from that to one level of abstraction even higher, which is okay, now how do you design any conversation? And the tricky thing with any conversation is that it’s not just that dynamic I mentioned where the user is coming inbound and you’re you’re running procedure.

It could be very open ended. Right? You’re you’re reaching out to them, and your job is to, like, charm them into activating their account or something. We’re already seeing some of that with some of our customers where we’re helping them with that. And I would say it’s still pretty early stages. That’s how you evolve from, you know, this the current framework into something that’s more.

Yeah. Those those have to be customized from customer to customers. You essentially have to give them the option to enforce these guardrails. And then our job is if the guardrails is set, then we need our infrastructure to be one where we’re heavily monitoring them and making sure it doesn’t happen. There’s there’s like a bunch of techniques you can use for this. You can have essentially a supervisor model that’s there when the core agent is running and its sole job is to check the guardrails and make sure that they’re all passing.

And if not, then it’ll instantly rewrite the response. There’s stuff like that that you can do, but it really depends on the company, right? Like if we work with financial services clients, they might say, we never want the AI to give financial advice, which is a very reasonable thing. And so that will be a guardrail that we have to put in. And, you know, if you’re in a retail space, probably you don’t care about that guardrail. And so you might have other ones since it has to be really customizable.

I would say from our experience so far, very few people will build it themselves. Robinhood is I mean, they have an exceptional engineering team, and they also have the different philosophies. And so I think there will be a couple pockets of players that are more of that profile that are happy to build it themselves, where they’re super tech heavy and generally on the sort of like the lower enterprise side of things. What we’ve seen in just the enterprise at large is that folks may have done some work internally.

Some people experiment with GPT somewhat, and they quickly realize that to actually get something production level, you need a lot more than that. You need essentially a whole framework for non technical users to come in and build and adjust things. You need alerting, monitoring, observability, the ability to run experiments, AB testing and stuff like that. Why build that internally? It’s just, it will take you way too long and it’s not really what your engineer should be doing. And so that’s generally what people come to and that’s when they decide to partner.

So I would say it’s not very common that we see people doing the full in house approach, but sometimes it’ll be a hybrid, right? It’s like, hey, there are certain things we want to own in house. And so then it’s our job to really flex around that and make sure that we can provide the support they need elsewhere.

Yeah. Roughly there. The lines get blurred because oftentimes what will happen is like, hey. I’m sitting down to do this project. Let me just have the agent take a stab at it first, and then you go in and, like, edit things and and move things around. So there’s a bit of, like, teamwork that involved, but, yeah, 50 seems about right. I don’t I don’t actually know these days. So

for sure. At our stage, we cannot get enough engineers right now. We’re hiring engineers so fast. In five years, I think we’ll have dramatically more engineers.

Well, actually, typically, I find we don’t compete with them too much. Like the the large companies, I think a few people who want to work there versus people that want to work at companies our stage are usually not the same folks. And so I don’t think there’s too much direct competition with those folks. But to your point, there’s a lot of companies hiring, right? So it is a big challenge to make sure that you you and the best folks. And so you just got to be clear about, you know, what what is the unique advantage of working at a place like us, right?

And I think we have a pretty unique culture. I think everyone here is is one where they’re opting into like, we’re in the office all the time and we’re working hard and that’s gonna lead to somewhere that’s gonna lead to me leapfrogging my career or financial outcomes in the future.

It’s just way more productive. My last company, were remote. So, because it was during COVID. And I think there’s pros and cons of that, but I think overall, just personality wise, just comes down to the founder’s personalities. Like, Ashwin and I are doing a lot of communication in person. It just allows the ideas to flow a lot faster.

Saturdays no. So Saturdays is essentially a spouse day because, you know, most of the team here is married, and so we we need to spend time with their families.

Yeah. So when I say team, it’s it’s more the the leadership team. I would say we all got married quite young. So I got married when I was 24, and Ashwin was pretty young as well. And our VP of engineering, VP of APM, they all married in the last year. So I don’t know. We just have a group of folks that got married quite young.

24?

Yeah.

Was it love at first sight?

Yeah. I mean, we met in college. So I was a sophomore. She was a junior, and first person I ever dated, and we just got along well.

Yeah. Probably be the two I mentioned. So Salesforce, we actually are pretty familiar with the Salesforce team. We have some folks on our team from Salesforce. And people like to dunk on the older, like, bigger players, but they’re also dealing with a completely different problem. They just have so much scale that you just have to build something that can take on that scale. And, yeah, I think they have a really smart team working on it. So we definitely respected that for them. I think that the competition there hasn’t heated up as much because they are more of a almost like a horizontal player.

Like, they they’re just trying to be horizontal off the bat, whereas we’re obviously starting from much more of a vertical approach.

It’s not really my place to say that. I would say what I’ve seen from the outside is that they may be getting pulled into too many different directions because they have too many customers. That could be one. And then there’s also the classic one where it’s just it’s just hard to move fast when you’re that big. There’s a lot of different considerations. There’s a lot of, like, risk profile type things that prevent you from watching things fast, and so it’s just hard.

share revenue directly, but last year we went from roughly 0 to 8 figures ARR, and we raised towards the end of that. So that was our Series B. And then Series C was, we announced it recently, but we closed it, It’s been a couple months at this point, but we’ve had tremendous growth this year.

I thought it

I mean, it’s those these things are all subjective. I will say we could have raised at a much higher valuation up to, like, 1.5 to two x. The reason why you don’t want to is that it creates couple weird dynamics. And so

So there’s a couple things just mentally, it just creates like, okay, well, now the goalpost is here. So all the wins you do have just feel a little bit diluted and you’re like, okay, well, now it’s cool. But you know, the goalpost is way over here. So and the whole team just feels a little bit demotivated. It’s like, okay, we’re working so hard and we are making progress, but everything’s been worked so high. The second thing is that it makes hiring a little bit weirder, because people will come in and like they’re getting their equity packages, but it’s like, okay, well, how long will it take you to get to this valuation?

Whereas I think that our current valuation, hopefully is not going to be that long. And so that’s the dynamic that you see. And then lastly, I would say it just really limits your optionality in the future. Like when you think about people that raise that huge valuations, I’ve known folks that have done this, you get to a point where you maybe are still doing well as business, but the markets change, and you’re just not able to raise at that price with some healthy multiple on it.

And so then it just feels like your business has lost momentum, even though maybe internally you feel like, oh, you know, we’ve been growing steadily. It’s just maybe not as fast as we could hope, but still at a fast rate. And then you just feel like a zombie company. So I think those are the main dangers.

Think there’s pretty rarely a enterprise market that is a winner take all market. Most of the winner take all markets are more marketplaces where there’s heavy network effects or, like, more prosumer type markets.

Well, Salesforce is a little bit different. There are some

No. That’s fair. I think our our space probably will have multiple winners. Hopefully, not too many and hopefully, or one of them. I think it’s it’s just hard to say, like, why there would be a a huge single winner here. The reason is that there’s not that many network effects from between customers, and then there’s also just different things that different companies want. So that’s the reality.

Oh, yeah. For sure. I mean, there’s switching costs with most enterprise software companies. And so that’s why the the sort of next few years matter so much is that a lot of people are evaluating this right now. And so you have to get in front of them and and make sure you showcase why you’re the best.

I mean, there’s some things you can port over, right? Like, you’re still going to have the conversations and you’re still maybe going to have these sort of outlines of your operating procedures and so on. But the thing that is unique to most platforms is that it’s almost like how people talk about system of record normally, where it’s like your stuff is saved in there. I mean, is a great system of record and they use this configuration language and stuff like that. So that you’ve already invested hundreds of hours to configure it.

So you’re like, okay, I’m locked in. That’s kind of a system record. I would say for AI solutions, it’s more of like a system of intelligence where you’re storing the business logic and you’re storing the way that your business works. Ideally, that updates over time as well, but that’s the thing that’s harder to port over. I will say what we’re noticing in the market though, is that people really care about not being locked in. People care that, hey, the solution that you have is more agile.

That’s why we’ve kind of sort of created our AOP system around this concept of like being a lot more transparent. So it doesn’t feel like, okay, I just bought this AI solution. And even though it’s working well, it feels like a black box. And so I feel like I’m just locked into it forever because like, I don’t even know how to how to move on this.

That one is hard to say. I don’t imagine it’ll be too different, but

Yeah. I think there’ll be more of a learning from examples that’ll happen. Because if you just think about how humans learn, if bring, like, a really good human support agent, they learn by shadowing and kinda seeing examples and so on. There’s probably gonna be more of that in the future. Right now, it’s prompting is the the best way.

I’ll give a sort of work related one and then more of a out there one. So the the work related one is I don’t know how hot of a take this is nowadays. I feel like more people are buying into this, but, Decagon, we we really value just, like, clock speed of really, really highly and clock speed. I’m just being like how fast your brain works and how fast you can learn. And this is across all functions, right? Obviously, it’s important for engineering, but for sales and for marketing and so on, like a bunch of our top sales reps.

They’re just like really smart people. And they’re just like figuring stuff out for first principles. Like, there’s go, go, go really aggressive. So clock speed matters a lot more to us. That matters a lot more than things like experience and necessarily being able to sell to this industry and so on. So

Oh, sorry. I misinterpreted it as he made the investment, but That’s

Interesting. I think we’re still not old enough to really know how some of these decisions have panned out. But I think mostly, most of our decisions at our stage are pretty reversible. So if we decided to, hey, we made a bad decision, we can reverse it. It could be higher as we’ve made in the past would be the main one.

By the way, I don’t I don’t actually don’t have like a bad relationship with any any of these folks that have left. I I still think quite highly of them. I think we just did not have a good enough understanding of our own culture. I don’t think we even crystallized this clock speed concept until recently. And so I think there were more cultural issues that we didn’t select for, and we were just selecting for, like, oh, okay. Like, yeah, maybe you know this industry well or or so on, stuff like that.

So

I would say that when I interview people, that’s the only thing I look for because the rest of the team does like the actual interview. My job is to I just have a conversation with them. I talk about the projects they’ve worked on in the past. I talk about how they make decisions. I ask them to explain stuff to me, maybe something that I don’t really know much about and it was something that they worked on. I mean, one part of it is like, how good are you articulating stuff?

But it’s just like, how fast can you think? Right? Like, I changed the topic and like ask about this other thing and that’s helpful. Yeah. It’s more of a general feeling. I wouldn’t claim that we have some perfect measure of clock speed. It’s just I think it’s become just a view that we hold pretty tightly.

It’s kind of a a a nuanced answer, but Ashwin had worked previously at Palantir. And both of our personalities, I think we really gravitated toward this, like, you know, forward deployed concept. The thing is, I think forward deployed and I, for some reason for deployed engineering is just so hot right now. Like people are just

Exactly. And so I think our mistake was we over indexed on that a little bit because I don’t think that concept necessarily makes sense in all use cases. And in our use case, the types of things that people care about across different customers is generally pretty consistent. So when that’s the case, you have way more advantage building a product first approach. And so over time, we kind of found this good middle ground where there are, of course, still forward deployed elements of what we do because we have large customers.

But I think within the space, we lean much more on the product approach and that allows us to scale a lot faster and allows us to enable our customers more as well, where they don’t have to feel like, hey. I need to message Decagon for everything.

the other one that I’ve been talking with people about is, right now a lot of people are really stressed or if you’re working really hard, you’re really stressed. Stress is generally seen as a very bad thing. That makes sense because you will like negatively impact your health and maybe shorten your lifespan and things like that. Generally, what people do after that is they try really hard to mitigate it and do all these like wellness things or like physical balance and things like that. And my view is that it actually does more harm than good.

And instead you should just embrace the stress and treat it as almost like an advantage in that. Also, there’s no stress in your life. You’re probably like, your life’s not as exciting. Right? So like you have to embrace a lot more.

Yeah. Sure. I think that is a bit of it as well. Yeah. Think my take is, like, if people are just, like, kinda over indexing all this stuff, right? It’s like, if you invest so much into LLS wellness stuff, it actually makes the stress worse. Because now there’s this huge juxtaposition between like relaxing and stress, well, instead just like embrace it, we’ve got to toughen up. And it’s like, it’s almost like a privilege that we’re in this situation where we can compete for all these big customers, and, we’re in such a fast moving industry.

It’s it’s quite exciting.

That’s one of my favorite businesses. He’s also one of my favorite founders. Like, the Honestly,

Exactly. The reason I like that I mean, we’ve taken a lot of inspiration from them, but that’s basically what I’m saying. Right? It’s that managing stress is not the thing that’s important. It’s like, hey. Is what I’m doing meaningful? And do I get joy out of these milestones? And the flip side of that is that at Decagon, we try really hard to celebrate wins. For us, our wins typically are related to growth milestones, but then we make sure that everyone that’s evolved is super celebrated At big milestones, the whole company comes together and we, we’re not afraid to invest in that.

I have a tendency to, basically, it’s very difficult for me to escape the low level details. It’s definitely not a very responsible use of my time at this point because, like, to lead the org, you generally have to make sure you’re spending enough time thinking about the high level strategic things. But I cannot avoid just like going each deal that is coming up. Like, I care a lot about this deal. I have to go like super in the weeds and the ref sometimes don’t like that either because it’s like, oh man, I’m getting scrutinized so much.

That is probably the main thing. And, you know, trying to to be a little bit more balanced there, but it’s just hard.

Oh, man. That is gonna be tough. And then our investors will be unhappy. I think our investors right now are good at very different things. But if I had to shout out one person, she’s a relatively small investor in Decagon, but we’ve been, or I personally and our team have been very, very impressed with, Anu Hariharan from Avra, especially on the go to market side. So, yeah, definitely shout out to Anu.

no. What I’m still here being asked, I think is just, I think the super broad questions are just like, it’s like, okay, where’s the company gonna be in five years? Like, that sort of stuff is way way too broad. It’s very tough to answer these questions. And, course, I can give our vision and roadmap. But

For us, it’s probably sales because go to market is such a big part of organizations like ours. But I do think in three to five years, people will figure out some things that where AI can be really useful. I think we’re still not using AI as much as we could on on the sales side. And I don’t think fundamentally, there should be a reason why, like, you know, companies like us are making such a big impact on the, you know, CX side, but sales for whatever reason is just like a constant thing that cannot AI cannot help with.

I do think people will find more creative ways there and that’ll really change the way that work is done. Even now, well, our sales team relies a lot on chat GPT to do, like, research and stuff like that. So that that’s, like, one simple use case, of course, but that’s, like, not that impactful in the grand scheme of things. So I do think in three to five years, there’ll be some big changes there.

Sure. Let’s do it.

Anthropic.

So one, I I I don’t think that market is a winner take all market, so I do think there’ll be many big players. OpenAI’s advantage is that it’s just on the consumer side, it’s just so dominant. Anthropic, I think there’s there’s a lot of interesting things that it’s doing that could surpass. So what I think their work on coding agents has been really impressive, and they’re definitely the ones furthest along there. I mean, that’s why Cursor uses Anthropic models. And also they just, I think they just, they have a good team, they’re marching forward and I don’t think that long term necessarily we’ll always see a 5x difference between OpenAI and Anthropic.

So I guess, yeah, maybe maybe if the question is more like what is spread trade, I think like that that multiple will shrink over time. I’m not necessarily that like, Anthropic will for sure exceed OpenAI.

The fact that there’s this narrative that AI is just going to transform every single use case. I think what we’ve seen is that most use cases have not been transformed by AI, especially if you think about enterprise use cases and and things like that is that, yeah, it’s just not quite good enough yet or the shape of the problem actually doesn’t really lend itself to AI. And that is way too overhyped right now, I would say.

The distribution of Salesforce, for sure. You guys 3,000 reps just instantly just going out there. Of course, that would be huge.

Yeah. I mean, hopefully, we’ll get there one day. That is a big advantage of of that. But, I mean

They’ll emerge over time. So, like, we now we we haven’t announced a lot of our big customers, but there’s lot of, you know, Fortune 500 customers, and I think they’re trying to expand beyond retail as well. The the approaches are just very different because of their leadership all coming from Salesforce. They’ve taken a very Salesforce esque approach where it’s, you know, the heavy configuration and sort of this longer lift to get to get going. And we’re more of the, you know, the product forward approach. And so I think those appeal to different types of customers.

I think customers that want to own stuff themselves and want to iterate, I think gravitate more towards our approach. But it’s also very early in the space. I do think that they have a ton of big advantages, and they’re also a smart team. And so they’ll make adjustments, they will make adjustments, we just keep going.

There’s a constant push pull of we have so many customers coming in. Okay. We have to like service those customers and maybe push out work that is more like longer lasting. But then it’s like, okay, yeah, but we have to do that work too. So there’s I think that’s like the constant like, okay, at what point do we invest? And that’s all kind of under the umbrella of all right, we hire, hire, hire, hire a lot of people. It’s a constant dynamic. It’s a it’s one obviously we’re fortunate to have, because we’re growing really fast.

But I think most companies in our position, there’s this constant tug of war between, alright, optimizing for the near term growth of getting these folks super successful, and then for the long term, making sure the product continues to improve and so on. And we’re lucky to have a lot of really great customers. And so we just naturally tend towards like, okay, we’re all in. We need to make sure all of our customers are really successful and they get up and running. And then we have our, you know, our our product folks are like, oh, but we need, you know, we need engineering resources to to do this other stuff.

Oh, the latter for sure. No question. I actually don’t think we there’s that much advantage in a lot of application layer sort of solutions from having a better model. There is some advantage for sure. Just

There’s some element of that. Like, I think the models are good enough for our use case. Yes. But I think most of the gains, the other way to put it in, most of the gains in what we do are from building around the models. Can you get it to be really accurate and, you know, orchestrate well? Can you have this, like, platform around it that makes it really easy to tell when stuff’s not so good and can be improved and automatically improves and so on? So I would choose having just unlimited access to the great engineers over probably most things right now.

My style is just naturally very intense. Same with Ashwin’s. And what that leads to is I think we have very much of this culture of it’s really geared towards winning for what we talked about before. And so far we have been winning. And so the team is quite happy and so on. I would say that there’s I think the best leaders have a bit of nuance there, whereas you kind of combine intensity with a bit more of the softness, I would say. And that’s something I’ve observed from some of the top founders out there.

Right now, we’re just so in the weeds. Like, every day is, like, so packed that we definitely lean much more on the intensity side. But I think over time, we would like to or I would like to be a bit more just wise about that sort of dynamic.

Yeah. Thank you so much for hosting me. A big fan of your podcast. I’m really glad we got to chat.