AI Agents Talking to AI Agents: Reinventing Commerce with Decagon CEO Jesse Zhang

0:05Today we're lucky to have with us at NoPriors Jesse Zhang. Jesse is the co-founder and CEO of Decagon, which provides customer service and other related AI for all sorts of different enterprises, including banks, telecom providers, airlines, and of course, many of the biggest and most important tech companies. Jesse Pryor started Lowkey, which was acquired by Niantic, and we're very excited to have him join us today on NoPryors. Jesse, thanks for joining us today on NoPryors. Thanks for having me. Can you tell us a little about Decagon and why you started the company, how you started it, how you all got going?

0:34Yeah, of course. So Decagon, for those who are not are really familiar with us, we're an AI customer service agent. And so you can kind of think of us, you know, if we're working with a large bank or airline or just people that have large contact volume, the AI's job is to, you know, have a very engaging and personalized conversation with the user and resolve it and, you know, save the company a bunch of money and, you know, ideally drive more revenue in the future because folks are more engaged. And as we've grown, it's kind of becoming more and more of a, you kind of think of like a conversational UI for the brand where It's how every user can interact with it.

1:07And we often use the term like concierge to describe this, but that's what we do. And you're working right now with some big banks or some of the world's biggest banks. You're working with airlines, telcos. Like you've actually gotten to very big customers very quickly. How did you go about doing that or how did it happen? Yeah. So, I mean, as you know, we started out mostly with the like digital native companies. A lot of startups do that. And digital natives, of course, are much more willing to try out startups. They can move faster. So that'd be like late stage tech companies and things like that.

1:32Yeah, like Rippling, Notion, folks like them, they were like great partners. And they also just helped us iterate on the product a lot. So that's where we started. As we've gone on, I think just naturally we're kind of pulled up market just because of the demand. And as you might imagine, that's where most of the large contact volumes are. So it just happened a lot faster than we thought. And I would say a lot of these enterprises also moved a lot faster than we would have expected. So that's why we ended up there. And that's one of the underappreciated things about AI traction is a lot of companies are willing to try things in a way they weren't willing to before because it's such a big technology shift.

2:09And so all these markets are kind of open now that weren't before that would be much harder to do. Yeah, I mean, another specific dynamic is that at the enterprise, it's becoming a lot more of a top-down motion. So, you know, in the past, any of these technologies could have been just like one team trying to vet it or decide it. But now it's like a AI transformation. And the C-suite, the board are all like very big on how do we adopt AI. And, you know, customer service is often one of the biggest areas or probably the most low-hanging fruit. So that's how these conversations have progressed. And how much of an impact are you having in terms of some of these teams?

2:44So I know that you're giving a lot of leverage to these customer service orgs. Are you making people two times more productive? Or I'm just sort of curious, is there a way to measure the outcome here? Yeah. In most of the large enterprises, the first thing they'll measure is just what is the, I guess, like efficiency that you're getting them. So whatever they're spending on their contact center or their operation, how much can you cut that down by? And we've done case studies now where, you know, folks have been able to cut that down by, you know, 60, 70%. Oh, wow. That's like a great success case, right?

3:11Because it's like a very clear business case. You can show it to everyone. And then the sort of secondary thing, oftentimes, you know, folks will even, you know, put this at the same level, if not higher, is just the customer satisfaction. So you need to measure that and make sure that your customers are, you know, having a good time and more engaged, if not just like more also just happy than previously. So you're basically providing these customer service AI agents slash workflows that help, I guess function 24-7 and multiple different languages out of the box? And do you basically do like a lot of integrations into what they are already providing or how do you tend to work with folks?

3:48Yeah, I think the way you should think about agents here are that it's more of a substitute for the mundane human labor. So whatever systems they're already using, generally an AI agent, at least when you first deploy, is not going to disrupt the tooling you currently have. So whatever CRM they're using, whatever telephony stack, we will just integrate with that. And then it's kind of doing all the tasks you would expect a human to do. And over time, that just continues scaling. And so one of the benefits of AI agents is that they're always on, you know, either awake 24-7. You don't have to train them, really.

4:23There's no churn. You can just, like, scale them out. And then you co-founded this with Ashwin. And you are both second-time founders. What made you decide to work on this problem in particular? Because I feel like many people's first company, they really focus just on the product and the technology. And then on your second company, you're often more likely to also focus on the customer side, the commerciality. Was that your story or were you always kind of more commercially focused in terms of how you thought about problems in the world to solve? Yeah, I mean, one of my, I guess, theses is that there is a lot of untapped potential and just like really strong technical folks in making them a bit more commercial.

4:56Because the types of problems on the go-to-market side, they're, I would say, generally a little bit more hairy. And so a lot of folks don't like the messiness and especially a lot of technical folks enjoy the engineering product problems more. But they're still kind of very interesting problems, very rewarding. And if you can do that well, that's how you get your company to grow a lot faster because you just do more sales. And at the end of the day, it's still problem solving. So, yeah, I mean, Oshawa and I, we're both technical backgrounds. We just got along very well. He's similar stages in life.

5:24We both started a company before, as you said. And the first time is when you kind of lack a little bit of the commercial sense. and you're just generally just trying to figure things out. It's very hard to build intuition of what is a good idea and what isn't. And so it is definitely easier the second time around. How do you think about how you hired or what sort of people you looked for for the team the first time around versus this time? Like, what are you optimizing for in the people that you bring on board in your second company? Yeah, I mean, we're a little fortunate now. I think we've built a bit of a brand around our talents.

5:56And I think we have a fairly interesting culture now. The way I would describe it is, yeah, we're generally just selecting for very smart people, first of all. I think we care more about that than direct experience and so on. I think early on experience is still quite important. I don't think we hired straight out of college for our first pretty large number of hires, but of course now we are. So you want a little bit of that blend, but the first thing we select for is just how smart you are. And that's worked out well for us. We apply that philosophy basically across the org. I think obviously engineering is generally easy to test for, but even on sales and marketing.

6:38So that's been a core part of our philosophy. The other piece is just we're in office. There's a lot of, I guess, fun news now. Companies work really hard. Yeah, yeah, sure. It's like the 996 culture and so on. I don't think we over-rotate on stuff like that. I think we're just looking for people where you can tell when you meet them that they really see this as like ideally like a like a they want it to be like a highlight of their career basically they want to put in the time and they want to be in a position where if they put in the time they'll get stuff out of it and they get to you know accelerate their career they get to work on like very very interesting problems so are you in office every day like five days a week in terms of when people are supposed to be in or yeah we're uh five days and then a lot of folks come in um on the weekends but it's not like a requirement yeah it makes Yeah, I mean, it definitely feels like you have sort of this hardworking culture people want to put in the time because, you know, it's interesting because if you look at professional athletes in training, they're always like, yeah, I train six, seven days a week.

7:35I work hard at my craft. And there was almost this period in Silicon Valley where people didn't want to say that. And I feel like with this wave of AI, suddenly it's come back that it's good to do that. You know, that's how you build a winning company and a winning culture. And so it seems like you all have kind of adopted that as how you approach things as well. Yeah, I think pretty much all the AI companies that are doing well have pretty heavy in-office cultures. It's just you get way more done, especially in the early stage. I think after a certain point of scale, like, yeah, you could definitely make the argument that it matters less.

8:07But as of right now, it matters a lot. Yeah, it also seems like there's certain roles that always have been remote, like throughout all of history, you know, in terms of certain sales roles or the like. Well, then really you're supposed to be at the customer side of your office, right? if you're doing some form of like field sales or the like. So it seems like a lot of people have sort of gone back to the pre-COVID era for the startups that seem to be working best, which I think is really interesting. And, you know, obviously things are working really well for you all. Yeah, exactly. How are you thinking about the main types of roles that you want to build out in the company now or things that you're hiring for or looking for?

8:38Right now we're kind of mostly building for scale. So what that means is, of course, we need to hire a lot more ICs. We're bringing in more kind of like leaders and adding a little bit more structure. I mean, interesting thing we're thinking about now is like a people function. Never really needed that. But, you know, we're approaching 200 people. It's you definitely need folks to be thinking about that full time. And it's more around like org design and what is the right way to structure our operating cadence between teams. We have an office now in New York. We're spinning one up in Europe. There's a lot more of those problems now.

9:11And so that's that's definitely something we're thinking about. I think that if you were to give founders advice around one thing that they should do that is against their instinct the first time they've scaled a company, what is that thing? Or how would you think about a big takeaway that you've had as you've gone from, okay, we have this nimble team that's grinding out a new product into, okay, we're scaling, things are working really well, we have product market fit, and we have to move as fast as possible. Is there a big mental transition that happens? Is there a specific tactic you'd suggest?

9:40So I would say for us, we kind of hit our stride fairly early in this company. So it didn't feel like there was a before and after. I would say when we were building, well, one, we stayed really close to the customer, which is always helpful. I think over time, the adjustment we are learning to make is thinking more like medium to long term versus short term. Because I think at the beginning, you have to think short term. you're just optimizing for closing the deal or closing a couple of customers but once you have your legs under you you both can think more long-term and also you have an obligation to because if you don't then eventually you get to a point where things really start breaking and you feel like oh man i should have you know scaled this better and so on so we're definitely in that journey right now we're trying to be as mindful of it as possible yeah maybe one related thing is that we do spend a good amount of time like studying uh sort of later stage teams that have done this well and there's obviously orgs that we admire where we who are some people you think have done it well i mean ramp comes to mind for sure databricks if you're thinking about a bit more like it's just like companies have just always executed well i think i think ali at databricks is one of the most impressive ceos yeah just in terms of like how he thinks about things and you know depth of reflection on different topics it's like really impressive yeah he's And actually, I would probably go far as to say he's my favorite CEO.

11:02And he's been very kind to us with his time. And that's another good example, honestly. It's like very strong technical folks that have, I think, also done very well applying that to commercial problems and execution. And that's definitely the DNA we want to build at Decagon. Do you screen for commerciality and the people who join? And if so, how can you do that? So say you have an engineer. Do you try to find people who are more commercial minded? Or do you think that self-selects a little culture? I don't think it's super important for every engineer in the company to be commercial-minded, for example.

11:32I think it's definitely very important for the founders, and then maybe the folks immediately around the founders. That's why I think generally when I talk to engineers that want to join startups, for example, and let's say they eventually want to start their own company, which is a very common profile, it's, in my opinion, it's way more useful to join somewhere where they've already kind of got the commercials figured out and you can actually see it in action and build that intuition than to join something pre-PMF. And I think that's like a very common misconception because it's like, oh, well, the smaller the team, like the closer I am to, you know, learning how to be a founder.

12:04But if you join a pre-PMF team and you never actually get to see the commercials in action, you're not really learning much. You're just kind of learning essentially what not to do. And unfortunately, the reality is that most companies don't hit that point. Our sort of like discussion we have with engineers these days is, hey, it's like, yeah, it's very important for you to join. And like, if you want to, you know, start your own company eventually, like, you know, Decon's like the golden age to do that because we have a lot of the basics figured out, but there's still so much that isn't figured out.

12:29And a lot of it is kind of very close to the commercials. Yeah, that makes sense. Yeah, I think a lot of the golden periods for many companies is between say 50 and 100 people up to 1 ,000, maybe 2 ,000 if the thing keeps going in terms of growth because that's the era where I think you see the most change. Although also going from 2 ,000 to 15 ,000 at Google, which is roughly when I was there, was also sort of this magical period of change. And so I guess it depends on the size of the market and the way the teams run and everything else. So, yeah, I guess also it seems like you can learn a lot more from success and from failure.

12:59And it sounds like in the context of Decagon, it's a really great moment to join because, you know, things are working and so people can learn different areas. Are there areas in particular that, you know, you'd really like to attract people? Like, is it international? Is it somewhere else? Yeah. I mean, the way I would think about that is like you're kind of if you're a founder, you're like training your own neural network. Right. And you need like positive examples and negative examples. Like for my first company, I mean, I started right after college. Like basically the first two years was just like negative examples.

13:26You're just like failing. And that's like helpful in some sense, because you can just kind of like brute force it and like try to like learn. But if you get some positive examples sprinkled in, your learning rate is just like way faster. Yeah, I think that's the misconception. So yeah, I mean, as we expand internationally, it's like, I mean, that's important too. I think an interesting thing with each new office is that you also have to kind of rethink the, like we worked really hard to build our current culture and SF office. And we've spent up New York. We're going to obviously send some folks out, but got to be mindful of that culture as well.

13:57Because once it's set, it's kind of like becomes its own living thing. Europe is a whole different thing because the culture over there is naturally a little bit different. And so you have to be a little bit mindful. It's also just also naturally more isolated. You have to serve wine at lunch and that kind of stuff. When you talk about having to shift the way that you think about things more towards medium and long-term planning, is that org design? Is that internationalization? Is that product roadmaps? Is it like, what is that? Is that capitalization? Like, I'm sort of curious, like what are the main components that you've had to start thinking longer term on?

14:29Yeah, it's probably to say it's more org design and product roadmap. Org design in terms of how you allocate resources. Because there are a lot of types of work that don't yield immediate returns. Like it's not going to close a customer for you. But if you don't do it, you will in six months, one, really regret it. And then two, you'll just be in a spot where it's like much harder to do that work. What's an example of that? Like just like core product work, right? Like, you know, there's a bunch of core product work that is important for closing customers in the future. It's not going to close any customers now.

15:00It'll probably still be fine for now. But you can definitely foresee that, okay, well, if you don't invest in this, then closing each incremental customer in the future will require the same level of work, if not more, because then you just have more overhead. And you want that to go down over time. And so that's the classic type of thing where you have to shift your mindset a bit. because I think in the early days, it's like really good to have a greedy mindset. It's just like, okay, I just really need to optimize for this one thing. Just get it over the line instead of just planning too long-term because if you do that, you could just end up burning a quarter and like not getting anywhere.

15:35And so I think over time, you have to make that switch. Did you set off to do customer service when you started Echagon or is that something that you all discovered early on as you were iterating on ideas or things like that? Oh, no, definitely did not come in with any preconceived notion. I had like a lot of empathy for the problem just from my first company. It was a consumer company. So we had a lot of users. But our general approach, kind of going back to the commercial side, was I think we're just a lot better at being commercial about this in the early days. And so we just talked to a lot of customers and had a very disciplined process of evaluating ideas.

16:11And yeah, it turns out that this has been one of the big use cases. What made you realize that this was the thing to do? The real answer is we just saw a lot of folks that were willing to pay us like, you know, six figure contracts, which at the time when you're at zero ARR, it's like, oh, wow, that's huge. And a lot of folks that were willing to, you know, do the same thing. And it was the only idea we really explored that really had that property where people were like, hey, yeah, like if you did this, I would literally pay you money because I can justify it. The sort of flip side of that at the time was more just, oh, well, this is such an obvious idea.

16:41Like, why do this? Because, you know, people would have thought of this before. But that's a whole other thing. I think like once you start doing anything, once you get into it, you understand there's way more nuance than the overall narratives. The sheer fact that people are willing to talk to us, like, you know, two people and willing to pay us money was signal enough that it was worth doing. I guess when I look at sort of the history of technology, anytime there's a big platform shift, the providers of the platform start to forward to integrate into the biggest applications on the platform.

17:08So an example of that would be after Microsoft launched its OS, it forward integrated in what became Office, right? Those were four separate companies doing PowerPoint and Excel and all this stuff. And then eventually they just subsumed the functionality of those things and cross-sold it as a bundle. And then that happened later with Google, where they started adding vertical searches for the biggest categories of search. If you think of that in the context of the foundation model providers like OpenAI or Anthropic, Anthropic is already providing cloud code. You know, they're already kind of forward integrating into different verticals.

17:38They mentioned financials as another area that they're moving into. do OpenAI famously tried to buy Windsurf and sort of enter coding more directly? Yeah. Do you think about that at all in the context of what you're doing, given just the size of the market and the velocity at which you're getting adoption? Yeah, I think it makes a lot of sense for the labs. I think OpenAI, for example, most of their revenue and most of their margin for sure is coming from ChatGPT and the application layer. You actually own the customer. You're kind of indexing more on the problem you're solving rather than the costs of your model.

18:10I mean, API business, for example, is, they're probably not expecting even to make that much money from that long-term and they probably see them more as a wedge. Some of those work out well, right? Like in other words, one could argue AWS and the cloud providers are good examples of what was perceived as a lower margin business that has enormous scale and can throw off a ton of cash. Yeah. And so, you know, these API driven businesses strike me as something similar. I'm just more curious, like, how do you think about defensibility relative to these things and, you know? Yeah, so I guess the point I'm trying to make because I think it makes a lot of sense for them to push into application layer.

18:41And I think they will. In terms of what applications, I mean, generally, they'll probably start with applications where it's more consumer prosumery because there's it's just more self-contained. It's easier to build the software on top. Long term, they may move into the more enterprise things. I don't think it's like super useful for applications like us to like, you know, spend a ton of thinking about like, you know, what the AI labs will do. I do think the more enterprise you are, the thicker the layer of software. It's not even just like, it's not even stuff related to the models. It's like, okay, how do you have observability and monitoring on all the conversations?

19:15How do you learn from the conversations? How do you really dissect the insights? How do you build a testing simulation suite for QA of the conversations? And there's just so much to build. That's what we're focused on right now. I think, yeah, it might make sense. And yeah, who knows, maybe one day we'll collaborate with the labs. We already have great relationships with the larger ones. But I think before they tackle our space, there will probably be other spaces they have to tackle first. Coding is probably one of them. I guess, John, I'm really to know, how do you think about differentiation?

19:42Like, what do you do uniquely? Or how do you think that you're built a mode over time? When we first started the company, this idea is very easy to grok, right? There's a lot of big platforms out there too. You know, like Salesforce, the Agent Force, and Google, and some of the more AI native players. What's worked for us so far is a couple of things. I think one, we kind of have a unique, we just have like a relatively young, intense team. And that has led itself to a couple of things. I think the biggest one is just speed. So we just been able to move really fast. And that shows itself in building the product and executing on the go-to-market side.

20:15And specifically in the products, I would say we've kind of differentiated ourselves in taking this approach of like, hey, this should be a very productized space. You should have an AI agent that's really easy for non-technical people to work with. and for them to build the agent, iterate on it, analyze it. And that's in pretty stark contrast to how the industry has always worked. If you think about the sales forces of the world and just the classic SaaS, it is a very more like a technical endeavor. You have to bring someone in to do the configuration. You have to have technical resources. As you scale, you can build something quite powerful, but it just becomes very slow and expensive to maintain because you use engineers to go through everything.

20:56And at the enterprise, there's so much complexity and nuance that you have to resolve that. So I think our view so far has been kind of different in that one of the things that LMs unlock is that you can really empower the non-technical business users. And that has, I would say, been pretty well received. I mean, different teams have different strategies, of course, but for the folks that we're working with, and especially as you go more upmarket, I think people really like that strategy. There are definitely some teams out there that are more engineering driven, which like the engineering team owns the entire customer service deployment, then maybe our current approach doesn't make as much sense.

21:33But I would say what we found is even when the engineering teams are very much involved, they don't necessarily want to be on the hook for every little change. And so in that case, we can work very well with them. And you have them still owning how does the AI agent interact with the systems and connect to APIs and so on. And then we allow them to offload sort of the logic building to the business users. So that's probably what's made us different so far. Again, obviously we respect like the sales forces of the world build amazing businesses, but we just don't think that's the right approach for the AI era.

22:00And then on our end, yeah, I think we're just, we really want to differentiate on execution. If you look at the big shift that's happening right now in AI, because of the capability set, we're basically moving from software as a service to basically some form of like labor or cognition as a service, right? And so you see that sometimes in the pricing models where people, instead of charging per seat, will maybe have some baseline platform fee, but then they'll charge utilization or other things because fundamentally it's almost like you're helping augment an agent versus just having a piece of software that they're living in or using.