Decagon’s Jesse Zhang: Closing Customer Service Doom Loops With AI Concierges

11:26Ashwin Sreenivas:It was mostly that. Like, I think it was mostly that we felt the pull. And even the early days when we were still like, we kind of were like, okay, cool, we'll help you with that. But, you know, I think that sounds too obvious. People would be following up with us on like, hey, have you guys reviewed this? Can you help? And you're like, no, not yet. And that was quite telling because we were just like two people. So they were willing to talk to two people, not only willing to talk to two people, but willing to go out of their way to follow up with two people. And so at that point, we're just like, okay, yeah, well, we should just listen to what we're hearing in the market instead of trying to intellectualize it ourselves.

11:57Jesse Zhang:Was it an idea that seemed exciting or worthy of your time, or did you have to kind of get hyped about customer service and sort of helping with these support centers?

12:06Ashwin Sreenivas:No, I think it's a very exciting area because it's very easily empathizable by literally everyone because everyone interacts with it, right? If it was a little bit more of a more abstract flow that we had never personally experienced ourselves, maybe. Although I would say I personally don't struggle with that, but I think maybe it'd be harder for us to hire. But I think because whenever we talk to new hires or customers, It's like, okay, yeah, we're deploying these AI agents in front of these contact centers so the customer experience is a lot better. It's like, okay, yeah, that makes sense. Because last week I was on a call and it sucked.

12:44Ashwin Sreenivas:If you guys could help that, that makes a lot of sense.

12:47Jesse Zhang:What made you confident that you guys would be able to build something that the incumbent players couldn't just tack on Gen.AI to do or other startups wouldn't just instantly do the same of as well? What was your edge or what made you think you guys had a unique point of view here?

13:03Ashwin Sreenivas:I don't think when you necessarily start that you have all that figured out. I mean, now we know, obviously, like we know our differentiation. But when you start, it's more of like everyone's kind of starting at the same point. So like everyone starts with like Gen. Like Gen. AI models became available to everyone at the same time. So everyone has the same starting point. We feel like Ashwin and I's strengths are that we're very strong at execution. We build really quickly. You just have to out-execute other people. Otherwise, you don't really deserve to break out. So that's kind of the mindset.

13:33Ashwin Sreenivas:It's like, okay, if this is a good market, then we're going to go and win the market just by overpowering people.

13:38Jesse Zhang:I guess part of why I ask is because, at least right now, it feels like a pretty obvious market where AI agents could be better than the phone robot decision trees that already were out there. And so it feels like, okay, it's almost like a starter gun went off. And it's like, go. We're a bunch of founders or companies pursued that possibility. And I guess I'm curious in the moment when you're starting Decagon, were you guys like, okay, this is going to be a crazy race or land grab and we just need to out hustle everyone else? Or were you even thinking about that at all?

14:11Ashwin Sreenivas:A little bit, but I don't think that's specific to like pretty much every startup like ever. Especially on these like big waves, which is where most of the startups get created. So like, you know, web to mobile. It's like, okay, clearly it's a big platform shift. And like a ton of these big companies were built like during the first few years of that or like on-prem to the cloud. and now it's like you know pre-gen AI to gen AI I think it's everyone knows that these are the opportunities where like an outsized number of like successful companies get created just because it's the platform shift I mean like I think everyone knows that and so yeah there is a sense of like okay like now's the time like we have to go and do it but if you don't feel like you have a competitive advantage versus other founders then like why are you starting a company in the first place so I don't think we've ever struggled with that I think Asha and I have always felt like we're quite strong.

14:57Ashwin Sreenivas:And that gives you enough confidence in the early days to be like, okay, well, in a level playing field, we can now execute.

15:04Jesse Zhang:But when you say we're quite strong, what do you think are the traits that allow someone to win a market like this? Do you guys just move faster? Are you more decisive? If you were to go one layer deeper, what do you think are the areas where you as a duo can out-hustle or out-win?

15:22Ashwin Sreenivas:There's basically three pillars that we think about. So one is on the classic engineering product, like how like your velocity of building. And we're strong engineers. Ashwin in particular is an extremely strong engineer. And then we've been able to assemble a team that's strong because of that, that can iterate on product very quickly. The second one is on the commercial side. I think both of us just like naturally like sales and we like learning about it. We like figuring out like how to improve sales. And I would say we're like quite a lot more commercial than the average like technical founder.

15:54Ashwin Sreenivas:And so So I think that's helped. And the third is, yeah, I think decisiveness is a good way to put it. It's just how aggressive and intense is the cadence that you're setting for the company. And I think if you can execute pretty well in all three of those, good things will happen. And it's just kind of how are you doing those three things relative to other founders in the same boat.

16:16Jesse Zhang:Memorable name helps too. How did you come up with Decagon?

16:20Ashwin Sreenivas:Yeah, thank you. We have so many people that's like, oh, this is such a cool word. It feels very strong. So my wife gave up the name. We were just kind of spitballing. We like the concept of like a 10X employee, which is what AI agents strive to be. And Decagon is like a 10-sided shape. It's a real word. We got the domain. It just kind of worked out.

16:42Jesse Zhang:That's awesome. Okay, so when did you guys really start to hit the ground running with this idea? And how did you score your first customers?

16:50Ashwin Sreenivas:So we first started selling, I would say, like probably early 2024. And we got the first customers just through some of those early discovery conversations, right? Like there were people that we knew or were friends of friends, and then we were able to go in and deploy. And I would say the sort of early year, the first year, I think we had a pretty good year, mostly because of almost like sales heroics, I would say. Like I can't really take credit for this because I think we got quite lucky with the initial team we hired. But I think in hindsight, we have some like absolutely monsters on our sales team where we hired them because we just thought they were like smart and like, you know, they seem like good people and we got good references on them.

17:31Ashwin Sreenivas:But that team has really set a culture of just like doing whatever it takes to win. And I think that that's been very helpful. So I'll give you an example. Like in the early days, we were still like a pretty small team. We barely had an office, et cetera. and uh you know we had customers that you know wanted to meet us etc and they were going to be you know in the city where you know the the salesperson on it couldn't be but it turned out that their their parents were there and so they had their parents host the customer at their childhood home and then he sent you know his sister some talking points oh my gosh and the customer loved that you know it's like oh wow like people are really going the extra mile uh we had another one where um we ran a deal and there was like an executive who was important to the deal, but he was super busy and like really wouldn't give us much attention.

18:18Ashwin Sreenivas:And so the salesperson kind of made up an excuse to like just go out and fly to that country and then like be at the office. Basically there were people we knew there and like we threw a series of just like, you know, asking around and like getting intros. We got time with that person, even though we really could not have it. We were just like emailing him over time. And he walked away from that conversation like really impressed like and like quite quite happy with us and so like it just really accelerated the deal you know it's just like stuff like that i think we got a little bit lucky and uh and that kind of snowballed so the the more customers you get like the easier the

18:52Jesse Zhang:next one becomes so you guys were good at selling were you also finding that you had a product that was winning head-to-head and was a better product and if so what had been the big unlock to make for

19:03Ashwin Sreenivas:a winning product yeah so i think the the core insight we had was this the you know speed of differentiation or speed of iteration that I talked about before. So if you think about like the previous generation, you have companies like Salesforce or the CEO of Salesforce actually started a company in our space called Sierra. And that approach is very successful, of course, for them, like massive businesses like Salesforce. It's like Salesforce, a little classic company that's like, okay, you're never going to replace them because, you know, they're, they're so entrenched. But that model is really based on this like lock-in.

19:35Ashwin Sreenivas:So it's like you're locking in people because the setup and the iteration is quite difficult. And if you feel like, oh, wow, we literally went through, you know, three to six months of configuration where we had to get Salesforce engineers to come in and get everything set up, like, of course, you're not going to want to replace that because it's like, oh, do we want to do that again? And so I think we've really just counterpositioned against that approach, where instead of this heavy iteration, we want to make it so that your cost and ownership of the product goes down over time. In fact, it's just very easy to scale.

20:06Ashwin Sreenivas:And that matters a lot because we're selling to like pretty large enterprises, like basically, you know, the global 2000. And within that, you know, people have a lot of complexity, you know, like they go live and there's like, you know, 100 other flows they want to add to the AI. And if each one of those is going to take heavy configuration, that's just not okay. And so our main differentiation is this format we released called the agent operating procedure in AOP. And so it's a way to teach the AI how you want it to behave and the procedure you want to follow in mostly natural language, while still being able to, you know, give it keywords to force it to do certain things in like, you know, regulated situations and so on.

20:42Ashwin Sreenivas:And so that format really empowers, you know, the teams that we work with so that the non-technical people can run quickly, the engineers can own, you know, how to access their systems. And so it kind of creates this harmony where you're able to just like run faster and faster. And that's the big differentiator for us.

20:59Jesse Zhang:As a busy founder at Upstarts, I don't get much free time. So magic my horror when, on a rare weekend getaway to a quiet cabin, I woke up to a message that someone was using my Upstarts card to make multiple thousand dollar transfers. Thankfully, Upstarts banks with Mercury, meaning I was able to block the transactions, freeze the card, and upgrade our multi-step approvals all from one dashboard, faster than it took to buy the cider donuts we ate to celebrate. You expect quick support from your other software tools. Why should banking be any different. Visit mercury.com to learn more and apply online in minutes.

21:35Jesse Zhang:Mercury is a fintech company, not an FDIC insured bank. Banking services provided through Choice Financial Group

21:41Ashwin Sreenivas:and Column NA, members FDIC.

21:44Jesse Zhang:And so you did mention, obviously, Brett Taylor launches Sierra. I wrote about you last year in the context of this horse race. And you actually said to me in the Upstarts article, you said that you didn't think that speed was a long term moat. And obviously speed has helped you guys expand, but how were you thinking about, okay, where is Decagon really going to stand out and be different over the long run?

22:06Ashwin Sreenivas:Yeah, so I mean, we have a lot of respect for the different players in our space. And so we're always trying to learn from like, what do people do well? I think what matters in the early days is of course speed. Because, you know, back to the analogy, if everyone's starting the same starting line, like in the early days, it's just like how fast can you go? But the speed should get you to somewhere that has, you know, more long-term advantages. And for us, that's in the product because obviously we really pride ourselves on our go-to-market execution, but I think I would put that in the category of like, you know, that's like an advantage you have, but that's not like a long-term differentiator.

22:38Ashwin Sreenivas:Most of the long-term differentiators are sort of product directions that you, a product like, it's kind of like a bet on this direction that either other people don't agree with or it just takes a long time to invest in it. And so for us, it's, again, it's this concept of the speed of iteration. It's like everyone will say like, of course, like, of course you want to have a product that allows you to iterate faster. but I think specifically the bet we've made that it is possible to do this in the product and if you can do it in the product then it's a huge win because you know everything's in the product it's very easy for the business team it's very easy for the engineers and I think for whatever reason other folks in our space have not really like committed to it the way we have or they seem to be moving slower in that area because they're kind of hedging across a bunch of different approaches, you know.

23:23Ashwin Sreenivas:And yeah, and so it's kind of the conviction in this approach. And we build conviction by working with our customers. And they choose us because they're like, okay, yeah, we believe in this, right? We work now with, you know, several of the big banks in the US and airlines and telecom. And the reason why they chose us over some of the other folks is that they're like, okay, yeah, we do agree with your approach. We do feel that it is, you know, in our best interest long term to be able to iterate fast in the product and that we can own it and we're not relying on you, et cetera, even though they do want us to help.

23:55Ashwin Sreenivas:So that's what's given us enough conviction in this direction.

23:58Jesse Zhang:Got it. Obviously, you could argue that having really strong competitors also validates the category, but would it be competing with the intercoms, the Sierras, the viable companies here where you feel like an upstart? Or would it be earlier on that you have maybe felt the most like you were punching above your weight as an entrepreneur here.

24:20Ashwin Sreenivas:Yeah, I think it would revolve around some of these big sales deals that we've done. What will often happen in our space is that, and I think this is, I'm sure it's not unique to us, but the deals that we run are, they're kind of like multi-level. It's not like, oh, here's a product, we get enough grassroots people using the product, and then that kind of bubbles up. It's more of a product-led approach. In our space, they're almost always top-down initiatives, where the execs are kind of pushing it and it's like the whole company has to be bought in almost to move forward with one of these solutions.

24:54Ashwin Sreenivas:And so we were working with a big fintech and it's kind of like the final two solutions that are competing against each other. And we're kind of winning over the working teams, both on the engineering side and the business team. But then we kind of hear that, It's, oh, we, like our competitor kind of got in touch with the execs and like, you know, they're good friends from before. And so it's just like, oh, well, like, you know, that always is like a big, big red flag. And so what we have to do is like at that point, you're, you know, it's almost just like, all right, well, this is going to be a battle.

25:32Ashwin Sreenivas:And we have to map out all the possible things that we can do to win. And so that's when like our team as well kind of really gets together. And so for me, it's like finding any possible way to get in front of these folks. And so we kind of map out, like, okay, who do we know that also knows them? You kind of draw out a very detailed chart of the entire organization and who are the people who are on our side and not. And then we're able to go in and just slowly start converting people. Because obviously you have to win the sort of boots on the ground. but then so it's almost like you have this tree and like you know you have like green dots and red dots and it's like you're kind of like working your way up and converting the red into green slowly and those are not easy and so sometimes people there could be a bias and so there's like almost like they don't even want to meet with you you know like you just find an excuse fly out and you're like hey you just happen to be happen to be here like me with other folks like and And that works.

26:36Ashwin Sreenivas:It's like they're not biased against you because they don't like you. It's more of like, oh, they're super busy. They have a relationship with someone. And so it's just like it's not really something that they're really thinking carefully about. But if you get in front of them and you kind of build a culture where your entire team can do that as well, that's how you kind of convert those into green dots.

26:59Jesse Zhang:So the lesson there is don't give up too easily and then get on that plane, basically.

27:04Ashwin Sreenivas:Get on the plane and just, there's almost a mode of like, you have to just do everything in your power to make something work because nothing is going to be easy. And that's, yeah, I think that really pays off.

27:16Jesse Zhang:Is there a code where basically it means clear Jesse's calendar, we're doing a scramble?

27:22Ashwin Sreenivas:Yes, unfortunately. The scrambles are like, I mean, yeah, I'm sure a lot of folks go through scrambles. It's like, ideally, you want to come out on the other end of a scramble with like something you learned so that you can improve your product and improve your processes so that the next time you're still going to scramble, but you're scrambling for something different. It's one of those things where if you look back over a year, it's like, oh, wow, like we had like very smooth growth. You just plot out like, you know, your like revenue or something. But then if you remember back to every week that happened, every week there was some insane thing where we had to go and push beyond our limits to get something done.

27:58Ashwin Sreenivas:So it could be closing a deal, saving a deal. It could be really trying to get a key hire and flying out to meet them. And so we have a team that's been built up. And the same thing happens to the team, by the way. I think one thing I've noticed is that in the early days, we have a lot of these beasts that can get a lot of stuff done. And they're very creative and very hardworking. and like that's still needed. But as you grow, you also need to start adding in like structure. Not only in the org, but just like, you know, the playbooks that you follow. Because then otherwise people feel like, oh, well, everyone's just like doing random stuff and there's no like cohesive strategy.

28:34Ashwin Sreenivas:But we never really needed that like last year because we were like a 10th size last year. So yeah, I would say like pretty frequently there's cases where we feel like we need to, we're like kind of pushing to get to the next level.

28:49Jesse Zhang:Was there a moment where you were like, okay, we have momentum though that I now feel good that at least existentially, this is going to be a good business if we execute well versus early on not knowing if it's going to be your next big company?

29:04Ashwin Sreenivas:I think we felt like product market fit in the first almost like three, four months. Okay, really fast. So we were kind of lucky. Yeah, definitely not the case in my first company. And the reason was, as we talked about before, there was just pull from the market. And we felt like there were people that just wanted this. And then we could feel just how large the market is. I mean, of course the market is so large because it's basically applicable to every business that has customers. That gave us enough confidence at that point.

29:30Jesse Zhang:Was there something that maybe in hindsight, if you hadn't done your previous startup as a first-time founder, might have tripped you up or been a challenge where, you know, the experience of having done it before maybe gave you the confidence or the knowledge to know what to do?

29:45Ashwin Sreenivas:I would say the biggest thing, which is kind of related to our discovery process to find the ideas, is that the thing you learn, at least I learned from my first company, that's impossible to learn without experiencing it is just like the intuition of like when to not spend time on a bad idea. Sometimes it's very difficult to just tell if an idea is bad at first value, face value. So you have to learn the mental processes for how do you validate if an idea is good. So in the context of is this something that's worth a bunch of engineering hours to work on? If you just went and asked someone, hey, would you use this?

30:20Ashwin Sreenivas:And they say yes. And like, cool, it's a good idea. You could just end up wasting a bunch of time.

30:24Jesse Zhang:Because people lie, right? I mean, you were talking about this earlier about you can survey the prospective customers. We did that with Upstarts. I feel like people are too nice. They will say they would pay for something that they would not pay for. And they almost don't want to hurt your feelings in a way where the honest feedback would be more valuable. So I guess you have to build your pattern recognition.

30:43Ashwin Sreenivas:Because they're like, oh, I'm on a call with this founder who's working really hard. I need to give them something. It's very similar to sales. In sales, people also tell you what you want to hear. And so you have to have a pretty robust system of digging into the truth and asking questions in a way where it's not really feasible for there to just be a fluffy answer. That's probably the biggest learning is how not to spend time. And I think as a result, we were a lot more efficient in the first year or two.

31:13Jesse Zhang:In such a competitive market, once a customer does start using Decagon, if they're happy, can you basically not worry about them? or do you have to always be one eye in the rearview mirror that they're going to get bombarded by a lot of other companies saying they're better than you guys, whether they are or not?

31:31Ashwin Sreenivas:No, I'm sure that's happening. I'm sure that's happening for literally every AI product. And so you have to build a product where if someone's happy with you, the next person has to be 10x better to displace you. And hopefully you've built enough in your head enough that it's impossible for someone to be 10x better. because it's always possible for someone to be like 10 better because if they're just like really customizing it etc but people are not really going to do a whole like rip and replace for a 10 better thing like it has to be multiple times better so decagon just has to be 10x better

32:06Jesse Zhang:than anyone it's replacing and then yeah within the band of anyone else who shows up as a new company saying like hey we're the the new decagon yeah i think that's true of most software which is

32:18Ashwin Sreenivas:why, by the way, that these like platform shifts matter so much. Like, why are we 10x better? I mean, it's not necessarily true that we're like 10x smarter than the previous like company that did. It's just like, okay, well, we have the benefit of being like Gen AI native, and everything we built is with these LLMs. Even if they were to tack on LLMs, they have a ton of baggage that they have to deal with in their previous product and their existing customer base. That makes it very difficult for them to keep pace. And so, you know, we've been able to produce something that is 10x better which is why there's a lot of folks replacing their previous solution or even if they didn't have a previous solution like replacing you know like not needing as many heavy operational resources right and you know obviously we can never just be like oh well you know we're done like it's hard for everyone to be 10x better than me like we're always on the lookout for like how the product can be improved and so if the best companies out there have like they start because they're 10x better and they get that they get off the ground but then they're able to just like keep their foot on the gas and build a culture of just constant innovation.

33:17Ashwin Sreenivas:So that is very difficult for someone else to be 10x better.

33:20Jesse Zhang:If everyone kind of started from this point of leveraging these models to sort of build products, what is sort of the most basic reason why Decagon isn't a wrapper that is sort of a commodity product over time and has its own differentiated technology built on top of the LLMs that are accessible to other companies.

33:43Ashwin Sreenivas:Yeah, it's funny. Like when we first started like early 2024, like the whole narrative was, all right, like all the wrappers are going to get absolutely owned. I think it turns out that a lot of the companies accruing the most value right now. If you think about our space, you think about coding or like legal or whatever, it's like, you could say everyone's kind of a wrapper around the models because we're in the application layer. But it just turns out the application layer is able to capture a lot of the value because we're solving the business problem. And so you can see that the labs are also moving into applications.

34:13Ashwin Sreenivas:For R-Space in particular, it is something that, of course, we have to be thinking about. But I think there's a couple of reasons why I don't think the labs or the underlying platforms will do something that really encroaches on this use case. The first reason is that most of the apps that they'll ship will be very broad but thin. It's like a product that, you know, people can just use, right? Like coding IDEs, et cetera. It's like they can do a ton of stuff. They're very powerful. But there's not that much like heavy software. Whereas where we're selling to, these are like big top-down enterprise initiatives.

34:49Ashwin Sreenivas:They're just like so much stuff you have to build out. Like someone would have to like really spend a ton of time thinking through like all the features. Like, you know, we have to be able to version the agent. We have to be able to run A-B tests. We have to be able to have alerting and observability and be able to kind of review the conversations afterwards. And so these are all things that is just like a thick software layer. And then on top of the thick software layer, there's like this last mile, which is like what we've been talking about so far, right? There's like the white glove implementation.

35:16Ashwin Sreenivas:And for better or worse, we have like a pretty built out, so a forward deployed or like post sales motion where, you know, we have product engineering and so on that like help with this process to build our AOPs, to help them set up the product. and that is a very difficult thing for us to build and operationalize, but that also means that this use case kind of requires that. And I think because of that, it's a little bit more protected or sort of orthogonal to what the platforms will be doing.

35:48Jesse Zhang:For people who are customers working with the brands that you help, do they basically interact with the product and know that they're interacting with Decagon or are you the behind-the-scenes helper who it's better if I don't realize that I'm talking to? What is sort of the end experience there today, and where do you want that to go?

36:07Ashwin Sreenivas:So most people know they're talking to an AI just because that's part of the... It's like you don't want to be misleading folks. But ideally, you get to a point where they're excited to be talking to an AI because they're like, okay, well, we know the AI is really good now, and because of products like Decagon, I actually enjoy interacting with it more because I just get my issues solved.

36:23Jesse Zhang:Because it's better than that phone call or whatever. It's better.

36:25Ashwin Sreenivas:And so we've had so many deployments where we track things like we did a deployment with like Oura Ring, you know, like they get a ton of praise for their AI on social media because the AI agent can do a lot of things. And so we saw that the customer NPS actually went up. And the other rate that we tracked is like what fraction of people just come in and there's like agent, agent, agent, like I don't want to bother with this, get me to a human. And that went from, it used to be 1 in 3 to 1 in 20. and because people, we kind of like earn back the trust of end consumers that are like are interacting with it, right?

36:59Ashwin Sreenivas:Because like we were talking about, we're so used to these bad experiences that you just want to barge in. But if you can immediately show that you're better, then it helps. So people know it's AI, whether they know it's Decagon specifically, that's not something we care that much about. Sometimes we have our branding on there. If it's a voice, it's just a voice, you know, so that part doesn't matter as much.

37:19Jesse Zhang:Within tech circles, long-running agents, whether it's Cloud Code or one of these competitors, have become a big topic and sort of what that unlocks. Has that been good for Decagon? And sort of what is that doing in your business right now?

37:31Ashwin Sreenivas:Yeah, so that's a big topic of discussion for us. And I think it's something the sort of businesses we work with are very excited about. And so the idea is that if you look at what these big AI labs are kind of pumping out now, it's a lot of these slow reasoning models. because one, they want to be good at coding, but you want to just be good at like solving tasks just in general. And the thing is, unfortunately, these models are not super relevant in like, let's say our voice agent because the voice agent latency matters so much. Like you have to be really responsive. You have to figure out like what the next step is and you have to be very conversational, right?

38:06Ashwin Sreenivas:And so you can't wait a minute for a response to come out. But the same concept applies in our product as well, where you could actually use these models behind the scenes and build like a slow reasoning agent that can just continuously be running there and like improving the main agent, right? And so what does that mean? It's like, it's monitoring all the conversations that come in, it's reading all those transcripts, it's reading all the documentation that you have, it's reading the AOPs, it's reading like even like the code that you have set up so that can connect to data. And it's taking all that together and constantly thinking about like, okay, what's going well, what's not going well?

38:39Ashwin Sreenivas:For the stuff that's not going well, like how could we improve it? Is that like writing a new AOP? Is that adjusting this one? Is that updating our knowledge base? and that's really powerful because it's almost like a it's not even just one colleague it's almost like you have 10 colleagues that are just like always on like they're monitoring your your your work and like improving it and these would be pretty manual processes before and so now you almost have these like two agents working together and so we have uh we just released a new product called duet that it's kind of like these two agents working together and the idea is that you can really leverage these slow reasoning agents to to help the the main one and i think it's definitely been our most successful product launch so far.

39:18Ashwin Sreenivas:And I think this is going to be a recurring theme throughout AI this entire year. Because if you look at the coding agents, that's what they're moving towards too. It's like, first you had your cursor doing like tab autocomplete. And then now a lot of the work is like, okay, you should just give it a task and like just goes and does the whole thing and maybe checks back in with you when it has questions. So that's the goal.

39:38Jesse Zhang:I've seen a couple of people on Twitter be like, agents mean anything now. Like the meaning of an AI agent has kind of been stretched in so many directions that like, what does it really mean? In your mind, what is the most important piece of an agent or what is least well understood about an agent capability or sort of what is an agent from a Decagon context?

39:56Ashwin Sreenivas:Yeah, I mean, for us, it's kind of, it's pretty general as well, right? It's like, you have an agent is language models or gen AI models like working together to accomplish a task. And in our primary agent, that task is having a conversation with the consumer. In the case of It's the task of constantly thinking through what can be improved and doing it. And in coding or in legal, you have similar agents that are just there to do work. And I would say that's a pretty good definition for an agent.

40:26Jesse Zhang:So we have these long-running supervisor agents or different agents, and then we have these more focused...

40:33Ashwin Sreenivas:Yeah, and they're both agents. They're just different types of agents. And eventually you could have a team, or some people use the term like swarm or something. It's like a bunch of agents working together, but each of them is accomplishing a task.

40:46Jesse Zhang:When you think about Decagon moving forward, what gets you the most excited and where should people be watching the next act for Decagon?

40:54Ashwin Sreenivas:Yeah, for us, it's really around this concierge concept. The reason why we use the concierge everywhere, I think it's just a very apt description for where our space is going. You start with customer service and being able to handle calls and chats, but that's just one form of conversation. And so there's a lot of other types of conversations that you could have. And I think we're just going to have a lot of modules eventually that can handle a lot of these different types of conversations. And again, they can be reactive conversations. So like, you know, the consumer is reaching out to you. They can be proactive or you're reaching out at the right time.

41:28Ashwin Sreenivas:They could be revenue generating or, you know, cost saving where, you know, you have a lot of people doing mundane tasks that they shouldn't be doing today. So I think that's what we're really excited about in the space to come. And I think when we talk to a lot of leaders, that's also what they're excited about. And so I think there's a very good alignment there. And so the way we've designed our entire products with AOPs, with Duet, is kind of in a way where it's agnostic of the type of conversation it's having. We just want to be very good at designing conversations and implementing conversations.

41:58Jesse Zhang:And when you think about a sort of always-on AI concierge experience, what does that look like in the future? Where does this get really exciting beyond the status quo we might already expect today?

42:07Ashwin Sreenivas:With a lot of the hype in this space, it's really centered around customer service, right? Because it's like inbound calls or chats and you have an existing contact center. And so the ROI is really easy for these leaders to map. And I think a lot of the folks in our space are kind of optimizing for that. The reason why we use the term concierge a lot is that we think it should be broader than that. It's like at the core, what we're building is an AI agent that can have conversations between a brand and the customer. and that doesn't have to be just you know reactive inquiries like you would be proactive as well so you could reach out at the right time we did a big deployment of this with with hertz where we because we were able to be proactive at the right time we actually decreased the number of unreturned cars by 25 because it's like you're reaching out to them on the last day like do you need a rental extension do you need help it's just like i think the the results really speak for

43:02Jesse Zhang:themselves how would they be reaching out by text or text or a call call okay and uh there's a lot

43:09Ashwin Sreenivas:of use cases like this right it could be like sales like not sales in the sense of like you know cold calling people but like um oh like someone's coming in and they're trying to sign up and like we actually because the ai agent and the concierge is there it increases the rate like a conversion or you can do like upsells and cross sells at the end of a conversation once you've solved someone's problem so that that's whole surface area of just like interactions and so So we almost think of it as like a, it's a new UI, right? In the same way that when people move from web to mobile, there's like a whole new UI that opened up where people can interact with you.

43:39Ashwin Sreenivas:Now this is like a completely new UI where it's conversational and people can interact with your agent as like a new way to just like get things done in your product. So that's what we're excited about.

43:50Jesse Zhang:You probably know all your customers by heart, but has there ever been a moment where you, without really thinking, ended up interacting with the Deccan products as a consumer and we're like, oh, here we are.

44:00Ashwin Sreenivas:Oh yeah. Um, yeah, some of the airlines, uh, we're not like, we're still not like a hundred percent rolled out globally now, but I fly so much these days where like, you know, I get to interact with their agents and like, it's, it's quite pleasant because, uh, I also know like how it works. And so it's, um, it's kind of like fun to see both sides of it. But yeah, I mean, I think that's back to the original point, right? I think one of the reasons why it's like gratifying to work on something like this is that, uh, it is something that experience you can experience in your everyday life. and it makes the storytelling a lot easier.

44:32Jesse Zhang:Awesome. Well, appreciate you coming on the show to tell us about it, Jesse. Awesome.

44:35Ashwin Sreenivas:Thanks for having me.