Bringing Code Gen Agents to the REST of the Enterprise With Serval’s Jake Stauch

0:09 >> Hey Jake, thanks so much for being here today. We're extremely excited to hear your story and where the company has come today and especially where it's going to go next. So, thanks so much for making the time. >> Thank you so much for having me. Great to be here. >> I guess our our beginning question is really less about the company more about you. We know you've done a bunch of different startups your whole life have worked on endure entrepreneurial endeavors. Maybe some of the inspiration was even from your mom. Maybe tell us more about you know your journey to starting Serval.

0:36 >> Yeah, absolutely. Well, if we go all the way back my mom was a professional pool player and ended up starting a pool magazine. So, I saw entrepreneurship up close and personal growing up. I I went to college and I was very interested in starting a company one day and I joined an entrepreneurship living group where we all lived like in a fraternity where everyone starting companies, which was super cool. Very intense because if you took a job >> Yeah.

0:58 >> they kicked you out. >> Wow. >> actually you got evicted from the building if you took a job. It's like, "Hey, >> of most places. >> Exactly, right? You think they'd want people employed, but no we were we were this campus group that just wanted to be full of people actually doing startups instead of just padding the resume. And so, yeah if you went and got a Google internship congratulations like pack your bags and move out. And that was a really good forcing function to focus my efforts and I started this company testing advertisements with brain scans.

1:25 >> Wow. >> And ended up dropping out of college to run that. That that eventually became a company that built a video games controlled by your brain >> Wow. >> to help kids with ADHD. >> Oh, very cool. And this was before Neuralink or what inspired you to >> Yeah. Yeah, actually it was it was before Neuralink non-invasive. We didn't drill into kids skulls. That would have probably worked better, but take them longer time to market. The idea was so I was testing advertising with brain scans and we could tell when people were paying attention to TV commercials.

1:54 >> Got you. >> And we had customers are were "Can I take this home and it on my kids?" >> Yeah. >> and see when they're paying attention to their homework. We thought, "Yeah, we could do that, but we could probably take that a step further and actually give them feedback on their brain activity while they're playing a video game and get them to improve it over time." Ended up proving that that was quite effective. So, ran that startup for about 7 years.

2:11 >> Wow. And then what happened after that? >> Yeah, so that startup was a rocket ship. early success, but eventually flatlined. Realized that, "Hey, medication actually is is free and pretty safe and pretty effective and that's tough to compete with." >> Sure. >> And so, ended up getting a call from someone actually my wife got a call from someone who was working at Verkada that she knew. And she was busy with another job and she said, "Hey, talk to Jake." And passed the phone to me and started learning about this startup making security cameras.

2:41 and decided, "You know what? Let me see what it looks like when a startup actually has product market fit." So, we were living in North Carolina at the time. We packed our bags. We got in the car. We drove cross country. and that was 7 years ago. I started at Verkada as an early product hire. >> Mhm. >> And got to ride a crazy rocket ship at that company and see them go from about 100 employees to 2,500 or so.

3:02 >> What sort of drew you to Verkada's mission? Or was it learning from the founders? Or was the role? Kind of what excited you most? >> Yeah, I think it was seeing a company that was working, honestly, because I had been grinding away at my startup for 7 years, never really finding product market fit. You know, having some success, you know, getting some traction, raising some money, but never really feeling that pull. >> Yeah. >> And watching what Verkada was doing, having started relatively recently compared to me running the startup for so long, I felt like, "I want to see what it looks like to actually have PMF, to see what it feels like and learn from that experience." I always knew I was going to go and start another company, >> Mhm.

3:41 >> but I didn't know what I didn't know. I didn't I never seen success, so it didn't feel like where was I wrong? I wasn't sure. And that's what attracted me to them. I at the time, honestly, was not super interested in physical security enterprise IT teams. But over time I got really into the technology. I liked that it was hardware, which is what I was coming from, and I liked the IT buyer. I kind of fell in love with the IT buyer because it's like a technical buyer that you really get to know in a B2B context, you get to build a relationship with and try to understand how to solve their problems. And cameras are cool. Like the it was cool to do early AI, pre-LLM AI, and and think about computer vision and how we can make that better. So, overall I over time fell in love with the product, but at the time it had nothing to do with the product. It was just I wanted to see a company work.

4:25 >> And I guess two big things happened there. You met Alex, your co-founder, and you learned about this IT buyer client feedback loop. Tell us more about maybe those two aspects of it. >> Exactly. So, early on in my journey at Verkada, I get paired with a new hire that's on the engineering side. So, he's a director engineering counterpart, I'm a director product, and we get paired up to go after some new markets. So, we're trying to figure out what are areas that Verkada can expand into, new product categories. And we start with one product that's local streaming of cameras, another product that is environmental sensing, and vape detectors for schools trying to catch kids vaping. And we end up building a lot of these products together, and we're always building with IT teams. IT is the buyer. We're talking to them trying to figure out what to build. And what's interesting is they're giving us a lot of insight into their problems that have nothing to do with anything that Verkada's building.

5:14 Obviously, it makes sense that IT does a lot more than physical security. IT is only relatively recently involved in the physical security world. And so, most of their world is outside of this domain. And so, when you ask these kind of classic customer discovery questions, you hear about all these other things that have nothing to do with what we were doing at Verkada. For a while we were just kind of dismissing those, trying to focus conversation back on things you can plug into the wall.

5:37 but over time we realized, "Hey, there might be something interesting here." And the area that we really started to see as a gap in the market was around IT automation. IT wants to automate all the stuff. They're really into technology. They've got all these parts of their job that are really boring and manual. And the tools they have audit for automation just weren't cutting it. And that became an area of focus for us over time and eventually became the foundation of Serval.

6:00 >> And was Alex also very entrepreneurial? Like you've been starting companies since before college. Was he also excited to go start something or was or did you have to convince >> working on startups in college. He was a CTO of a startup before joining Verkada. So we both joined Verkada coming off the backs of startups that didn't have the outcomes we were hoping for. >> Yeah. >> And I think both joined Verkada thinking this is a good place to be while I figure out what comes next. I think we both thought we'd stay there for a short period of time and then immediately go back into founding something, but we we really got attached to what the work we were doing and got to learn a ton from being there. And that became a great launching point to go into Serval.

6:36 >> And how did you guys decide to make the leap and join Serval? I know you were a new dad at the time and you know, we got lucky enough at General Catalyst for you coming in the office and you were brainstorming different ideas and you and Alex had teamed up, but what gave you the gumption to really make that jump especially at such a crossroads in your personal life? >> Yeah, I think it's just something you feel compelled to do. I I I also dropped out of college to start a company. So it felt like the only thing I could imagine myself doing is running a company and that was always the plan. And it was it was mostly the timing was not great.

7:06 I was a I had a newborn. Alex had a 1-year-old. >> All right. >> but it just felt like especially with AI starting to to really feel like it was reaching the point and this is late 2023, early 2024. Reaching the point of productization where you could really imagine building enterprise applications. I think all these forces were coming together. And it was actually my paternity leave that kind of gave me the space to really start ideating and thinking through this. And that's actually when we met.

7:32 and and when we started first really talking about Serval was when I was on pat leave and trying to figure out, hey, like what am I going to do next? And and luckily it's worked out so far. >> For a lot of folks who maybe want to be founders one day, what sort of customer discovery did you do before, you know, raising money and launching the product, building it? Did you talk to a lot of the same people from Verkada or how did you kind of verify that this was the right idea? And maybe tell us a bit about the early days where >> Yeah.

7:59 >> building the tech wasn't super easy at least at the time. >> Yeah, absolutely. So, we did a ton of discovery outside of Verkada. I mean, obviously we had a lot of background of talking to customers at Verkada, but it was in a very different context. So, we kind of started from scratch, non-Verkada customers, trying to get a blank slate view of what do people want. And I struggled with the customer discovery process because IT is not a complaining bunch.

8:20 >> Right. >> And so, you want to ask them these questions of like, "What is your biggest pain point? Like, what's keeping you up at night?" And you don't really get satisfying answers. They're like, "I don't know, maybe a cybersecurity breach." You know, nothing that I can latch onto and and really dig into. And so, I pivoted my approach because I realized they're not going to call out and say, "I've got this burning pain." >> Right. Build me X.

8:38 >> IT solves things for themselves. So, you know, they're a they're a capable bunch. And so, instead I started asking this question of, "Okay, but if you could hire somebody to sit next to you and do work for you, what would you give them?" And framing it in that way was such an unlock because then they start talking. Because they've got a bunch of work that is not their favorite work to do, but they don't think of it as a pain. It's just part of the job.

9:00 And so, then they start thinking about, "Well, I'd probably have them like figure out this automation I'm trying to work on for this onboarding use case and I would probably give them all like all the help desk requests that are coming through." And what starts >> to emerge is a more consistent picture >> of no, automation is a pain point and it's actually the building of the automation that's the pain point. Obviously, there are plenty of automation tools out there. There are all kinds of drag-and-drop workflow builders, but they didn't have the time to go and build this stuff.

9:25 >> Yeah. >> And so, the core idea of Servo really became, "What if we could automate the automation?" And what would that actually allow? And it turns out what that allows is an employee support experience. >> Got you. >> Because employee support is an automation problem because you just want to automate all of the employees requests. And so if you solve automation, you automate the automation, then you can actually build a better support experience for employees. So that was the >> faster for those employees.

9:48 >> Exactly. Exactly. Because if you're doing a drag-and-drop workflow builder for a thousand different employee requests, like you'll never get there. And then all everything will change over time. >> What's the first product to build though cuz it's such a broad surface area to go after, you know, ServiceNow or someone like that in the ITSM. How do you figure out where to start? What was the first sort of point product? >> Yeah, I mean I think one is knowing where you're going was really important.

10:08 So we we wanted to have a target, a company to go after, and ServiceNow was was the obvious one in the space with workflow automation and and employee service enterprise service management. but yeah, where do you start? What we want to start with is what is the biggest bet we're making? What is the hard part? And build the hardest thing first to prove that that could be done. And our idea was, hey, if if automation is really what we're going after, what do we think is the gap and why IT teams aren't automating more things?

10:36 >> Yeah. >> The gap is that it takes so much longer to build the automation than than to do the manual task. >> Right. >> They're just defaulting to do the manual task. >> Gotcha. >> What if you could flip it? So you could just describe whatever it is you're trying to automate cuz usually you can describe an automation in one sentence, two sentences, and then it's going and building that that becomes the hard part. It's like what if you could just start there and let the automation build itself? That felt like the fundamental thing we needed to prove because if we're going and building another drag-and-drop workflow builder, we're not going to be able to compete in the market. We have to fundamentally reimagine what this looks like. And so that's what we decided to build first is code gen agent that could take your natural language description of an automation and then turn that into the code that will execute that automation.

11:17 And in 2024 in April, that did not work. Right? We're still kind of in like the autocomplete days of code gen. >> Was that because the underlying models weren't really there yet? Or was it that you had to keep fine-tuning your own model? What kind of was the blocker? >> The underlying >> were not there. I think this is ChatGPT 4 just come out and we were just running into a lot of different issues. And but what we saw was it kind of worked.

11:41 >> Right. >> If you gave it an example of here's the API, here's an example of workflow you might want to use and gave it all this additional context and kind of like cheated a little bit. >> squint and see it getting there. >> and see it getting there and I think from first principles it felt like coding is definitely going to be an area that these models are going to be very good at. It's a it's a verifiable domain and so you can imagine there being one a ton of data and two the ability to have synthetic data and automated checking of of results. And so it felt like this is an area that's going to get better especially for what we're building which are these kind of procedural scripts, automations at the time. And so we had a lot of confidence to keep going. And that's kind of what we built first was was hey, can we prove to ourselves that this is going to work and we kind of sort of made it work for a set of use cases.

12:27 And then you can build all the other stuff that is hard but is known to be doable like building a great ticketing platform, building an access management system for all the access requests, all the other things that you have to build. but we had to prove the hard thing first. >> And the underlying models getting better and better. At first it was existential. They had to get better for it to really work. Now as they get better it flows through and you see all these product improvements.

12:52 Is there some sort of threat though as well that companies we back such as Anthropic as their models get better they can do everything that Serval or a service now does or is that the wrong wrong kind of logic and reasoning behind it? >> Yeah, I mean it certainly you never want to write it off because these are teams full of incredibly bright talented engineers with a lot of capital and success at their disposal. But I think it comes down to focus. We are really focused on solving a very particular problem for enterprises and going it it is a large problem but in the scale of what Anthropic's doing these days it's a relatively small problem which is enterprise service management. And just as a sense of scale Anthropic in six weeks, added more ARR than ServiceNow did in their entire 20-year existence.

13:32 And so when you think about it that way, does it really make sense for Anthropic to put their best and brightest people towards a target that is not going to give them anything that is exciting in the very near future? >> Yeah. >> And so I feel like there's a massive market for us to go after. Obviously ServiceNow is one of the largest enterprise software companies of all time. but it's still not worth really diverting resources. And so it comes down to focus. And I think this is generally why startups can compete with incumbents all the time is the focus and the speed towards a very particular domain. And and that's what we think we have going for us. And I think so much of our product is not just the AI as well. This is not a wrapper on top of the models. This is a platform IT service management enterprise service management that existed long before we had AI. And it is enhanced by AI, but you could buy Serval and not have any AI capabilities.

14:21 >> Right. >> And I think that's fundamentally different from some of the other companies that I think are more in the kind of path of of the hyperscalers. >> I also see sometimes you can take an Anthropic or OpenAI model and use AI to automate and maybe cut some tasks, cut some costs on a one-off basis. But to really reinvent a product for the end customer requires a lot more than just the underlying model by saying, "Hey, your fundamental business could look different, could be different." You have to understand that customer and their workflows and their data in a way to think about how do we expand the TAM or expand the product versus sometimes just taking an open-source model or an Anthropic or OpenAI is more about an automation solution that's solely around cutting costs. So >> Yeah, for sure. And the long tail of the problems you have to solve is so long.

15:04 But I think yes, you could have a chatbot that fetches some information from a knowledge base and says, "Here's the Wi-Fi password." No doubt Anthropic, OpenAI, great solutions for those kinds of problems. But what about an approval to an access request to a sensitive engineering resource? And that approval is a manager. And that manager might go on vacation. And we have to sync in his calendar and figure out well, who has to approve that access to that sensitive resource when the manager's on vacation, and how do you delegate those permissions, and who decides who can delegate permissions? Like that long tail for just that one use case, think about that across the enterprise. That is something we embrace. We say we love it. We're going to do all of it. That doesn't make sense for a lot of these other AI companies to really focus on building all that software.

15:47 >> That makes a ton of sense. Obviously, ServiceNow has been thinking about this for decades plus, but has a major disadvantage that trying to put an agentic system on top of, you know, decade plus software really is difficult to do. Maybe talk a little bit about, you know, how your product is better than ServiceNow, how you've even been replacing ServiceNow with some customers, and kind of what the value is in their mind. Is it the cost cutting?

16:09 Is it the ability to do more? Is it the ability to improve their clients under overall revenue and profiles? Like what's really driving folks to adopt Serval over a well-known incumbent? >> Yeah. I do have to acknowledge that ServiceNow did so much right. I mean, they they they won the market in in a huge way. And so, you want to understand what they did right and actually do better at the things they do really well instead of just exploiting the weaknesses. I think you obviously have to do both, but understanding what they did well is really important. And what they did well is this configurability and customization. They could build this flexible workflow system for any company and go in any company and map their workflows and build a very custom experience for any organization, no matter how crazy long tail their use cases are. And so, we have to do that, and we have to do that better. And so, where Serval differentiates is the flexibility of this workflow engine, which is all code based on the hood. So, it's this core decision we made to essentially be a code gen application that builds all of these workflows, all these tools for not just resolving help desk requests, but for onboarding and offboarding and fetching information from the knowledge base, and updating asset tables and databases, and having code gen be that central primitive allows end users to go and build all these crazy automations instantly.

17:26 >> Yeah. >> And not have to go and contract with consultants or in-house service now developers to go and piece together a custom application, but actually build that within Serval. So, I think starting as an AI AI native company built around code gen is what allows us to >> Mhm. >> build on top of a lot of good ideas that Serval now had, but do it so much faster in a more flexible and more future-proof way because this is where everything's going. It's code gen application experiences.

17:50 >> So, your end customers really want to be able to do more work. It's not a cost-saving sort of calculus for them. It's a we can do all of these things with Serval that's much harder to do and much more expensive to do with a competitor's. >> Yeah, IT is such an untapped resource. You have these like really motivated technical folks in your organization and they're running around resetting passwords and doing this click ups and they want to build.

18:12 >> Yeah. >> And they now have the tools to go and build things. And then that's I think one of the most exciting things about Serval is that we're giving them the ability to go and build and do the things in technology that they've always wanted to, but they've been held back by just the the drudgery of the role. >> Very interesting. And in terms of team and your team, you know, a lot of folks say, well, applied AI is sort of easy because the models are getting better, but at least what we see at General Catalyst with their AI enabled roll-ups and with the many of the companies that we back is it's actually getting more complicated which model to use.

18:42 Anthropic, OpenAI, Google all have great models, but also the open source models are only four or five months behind. There's all kinds of different toolings, software, agentic workflows to put on top. If anything, being an applied AI engineer is more complicated than ever. You've managed to attract an amazing group of folks from Verkada and Rippling and all these great places. maybe talk to us a little bit about the culture you have set at Serval and how you're attracting these applied AI engineers who have offers at so many different places including the labs or other competitors.

19:12 >> Yeah, I think part of it is is just creating an incredible place to work. An incredible place to work is is incredible colleagues. Like that's what we believe is if you start with a core group of people that you want to spend your day with, not just because they're they're nice and great people, which they are, but because they're some of the most talented hard working people imaginable. You come into our office, you meet our team, and you feel like I want to be here. This is the group I want to be a part of. And so being laser focused on that culture of building that core group and then having that core group recruit people beyond their core group that match those same values and qualities, that was really important for us and that that becomes kind of a gravity where they pull in more and more folks that have those same attributes. And and now we're we're expanding beyond our network into into like the general market, but still I think I I I bring it all back to like the ability to bring a initial group of founding engineers together that kind of embodied our values that are incredibly hard working, incredibly talented, incredibly incredibly gifted engineers.

20:11 >> And you have a program actually for future founders, which I thought was really unique and interesting, where they go and become forward deployed engineers and really get their hands messy with applied AI. Maybe tell us a little bit about how did you come up with that idea and what's the pitch to anybody watching this who maybe wants to be a future founder, but the first stop is Cerval. >> Yeah, absolutely. Well, I think what all of us in AI are discovering right now is the models are really good.

20:34 >> Right. >> But then going into a large enterprise and transforming them with AI is still really hard. And that gap between what the models can do and implementation and transformation of large enterprise, that is the gap that Cerval is trying to solve and that is the gap that forward deployed engineers, deployment strategists are are helping to fill. Because what ends up happening is you want to talk to this customer, understand all the requirements, and then translate what the product can do into those requirements and probably build a little bit of additional product capability to make that possible. If you think about the person who's really good at that, talking to customers, understanding what they need, building what they need, then trying to sell it to them.

21:10 >> That's a founder. And so when we're trying to recruit for this role, what we're really thinking of is like, well, what are all these qualities mean? These are These are all qualities of founders. And so we have built out this program for future and often former founders to go in and build up their skills by doing exactly what it takes to be a founder, which is talk to customers, build what they want, and then sell it to them. And it's been tremendously successful. I mean, we we've scaled the program really just over the past few months since we started the program. And I think that we're delivering so much value to the future founders because they're getting to explore these big large enterprises and work not just with IT teams, but HR, finance, legal, all these operations teams, figure out all their different problems, and then go and build and deploy AI solutions, then figure out how to sell and scale that across the organization. So they get to touch all these different domains all through this lens of enterprise automation, and that I think is such a great general training set for whatever product they want to build next. it's not limited to just like a very very niche kind of IT help desk use case. It's It's They're seeing problems across these large organizations.

22:15 >> I think especially speaking with the clients is so key cuz a lot of people think you build a product and then, you know, the clients come. And we've both, you know, built businesses before, some that worked, some that haven't. And being embedded with those actual clients is very much the most important part of being a founder, figuring out what they need and and building against that. So >> Building against it, building the wrong thing a lot of times. You know, I think that's such an important lesson that you don't get a lot of shots at when you're when you're just going it alone because a lot of times people never even buy the thing that you're trying to sell them.

22:44 But it But getting the opportunity to buy, build, and ship the wrong thing sometimes, fix it, build it again, you get so many repetitions doing this work at a company like Serval that then when you go and you build your own thing in a couple years, you're going to have all these learnings of like, oh, I know I know like this seems like the right thing to to but it's actually not. >> Now, I definitely want to give you the opportunity to brag a bit about some of the amazing customers you have. I feel like all of the fast-growing AI companies, Mercor, Together, Perplexity, Ramp, so many of them already work with you guys, even though you're a 2-year-old company. Maybe talk a little bit about that journey, and I know Tatiana, your wife, has been a huge part of that as well, but how have you so quickly got so many of these customers, and then we can spend some time also on maybe what's next in in terms of scaling up?

23:27 >> Yeah, absolutely. So, it was hard to get the first one. Perplexity was the first customer that actually closed that we had revenue from, and it was over a year from starting the company that to to get the cus- to get that customer because we are replacing an existing legacy system. We have to have so much service area, and startups don't need an ITSM. Small startups don't need an ITSM. So, you have to be probably a couple hundred employees before you bring a tool like Servalent.

23:50 >> Right. >> And so, it took us a while to get to that level of product maturity. Then, you have barely enough product maturity to sell, and that means that the only companies that are really going to value your approach where you're basically cutting off a bunch of legacy features, but then adding some cool new stuff, are going to be these fast-moving tech companies that one, believe you're going to build and ship all the stuff they want that are that's missing, and two, appreciate the approach that's kind of unique and and novel here. And so, that ends up being fast-moving, great startups like Perplexity and Together AI, and and Mercor and and Ramp and and Clay and and other great organizations, but what took us by surprise is how quickly that expanded into larger and larger accounts.

24:31 >> Right. >> You know, servicing accounts with over 100,000 employees. You know, Fortune 20 accounts. >> Yeah. >> that happened very, very early in our life cycle, and I think one of the interesting thing is that there wasn't a big difference between what they were looking for and what a smaller tech AI-native startup was looking for. And that's a that plays to our advantage because we can build one product that actually serves both of those markets quite well. Now, the implementation, the support, like all of the other things about the deals are very, very different, but the fundamental problems they're trying to solve are actually remarkably similar.

25:05 And so, we're so grateful for the early AI native startups and and even more recently, you know, companies like Harvey and Lagora and and folks like that. But, where I see this business increasingly going is is outside of tech in categories that are, you know, you wouldn't think of as early adopters like Blue Diamond Growers, the almond company, one of our customers. And everyone has the same problem. Because you think about employee support, everyone needs to support their employees and even if you limit it to IT support, every company that has computers is going to need an IT support function. So, the market is literally every single company on Earth.

25:42 And we're seeing a huge opportunity in in outside of tech and in these larger legacy organizations. >> Yeah, it's been incredible across the board. It's media, government, large corporations, manufacturing. It seems to have hit all of them. I think maybe one part that resonates is IT is often thought of as a cost center and the tools you are giving to people to build actually can help them think about their product, their revenue, and instead of being a cost center, IT can now be a really important source of information and productivity for companies. And that seems to be something that resonates in all sorts of companies, not just in one particular vertical like tech.

26:18 >> Yeah, absolutely. We're enabling IT to become builders and they're increasingly becoming builders on the team, not a cost, but actually a producer of all this value for the rest of the company because you come in through IT and then IT can bring these tools to HR into finance and legal and all these other departments and then everyone is building on a tool like Serval and and IT is kind of the start of it all. >> And do you see the product evolution going that direction as well, where you address IT needs, but now you expand and you're going to human resources or other things that are very close to it or will the focus remain more IT?

26:48 >> I think we love starting with IT when we engage with an account just because a lot of times they're going to hold the keys and the credentials on how do we get access to all these systems, but we very quickly even today expand into HR, finance, legal, all these teams. In fact, the average company on Serval has 13 departments. >> Wow. >> on Serval. And so IT actually ends up being a minority of the departments when you think about it that way. but IT is often the best landing spot to spread from there and it makes sense cuz they they own the keys to a lot of technologies and they're increasingly you know, as part of this building mentality of bringing in a lot of AI tools.

27:20 where we see the product also going is a push more and more towards autonomy. This idea of instead of telling the system what I want, >> Right. >> the system figuring out what my problems are automatically and building solutions for my problems. The best way to support employees is not to automate their support request, it's to prevent their support request. It's to fix their problem before they know they have a problem. >> What's an example of that that you would point to when you're selling into customers?

27:45 >> Yep, say you get promoted onto a team and you have a different set of permissions than everyone else on the team, right? You don't have access to all these other applications and tools or you have very different set of roles than the other people on your team. Chances are you are going to request access to all of those things. >> I hope so, yeah. >> right. And and then you're going to have to wait for that access. It's going to go through approvals and everything else. Well, wouldn't it be cool is we could have a scanner that detects like, "Hey, this person doesn't have the tools that they need to do their job. Why don't we just give them access before they need to tell us that that's missing?"

28:19 >> Mhm. >> And from a security perspective, it could also work in the reverse of, "Hey, this person has access to all these things and no one else in their department at all has access. Maybe we want to like trim that down or maybe they haven't used this access in a long time. So from a security and cost perspective, let's cut that off." And so just this background agent concept of how can we discover all these problems not just in the IT org, but in all these different organizations, and then build and deploy the tools that solve those problems. Again, building Serval around this code gen primitive makes us uniquely qualified to do that because we can build the tools that investigate the problems, build the tools that then remediate the problems.

28:57 >> And I feel like Serval is now a really well-known brand here in San Francisco, especially among fast-growing tech companies. If you're optimizing for AI, you're probably already using it. What sort of the plan to get the word out more nationally and even internationally? Like, how do you think of building a go-to-market? You've done amazing job building this team and product in 2 years. And I think you have as aggressive a plan to build the go-to-market in the next 2 years, you know, starting probably a few months back. what sort of the plan around?

29:23 How do you hire folks? What are you looking for on the sales side? You know, how does Serval become a household name? >> Yeah. We're We are not we are not AI-pilled when it comes to building a sales team. We are going to build a massive sales team. We're not replacing them with with AI SDRs. We are we are building a massive sales team, many of whom in person, many of whom remote for our strategic accounts. And and that is going to be the kind of tip of the spear going into these accounts that are that are currently dominated by legacy competitors.

29:51 and then on the marketing side, we've been a big fan of events. We're going to do over 100 events this year. >> Wow. >> between conferences and and kind of curated experiences and and all these other dinners and and things that you can imagine. And I think especially for these strategic accounts, which is our focus, bringing people together and letting them talk with us about their problems and talk with them about what we're seeing and what where the market's going. And as much as possible, getting them hands-on with the product. I love going to a conference just like having people try the product live.

30:19 I think that's the way to get the message out is just be in all the places where our buyers are. And I think increasingly as we go out market, those buyers do, you know, they do show up to these things. And we can we can reach a lot of them at the same time. >> And then I have to ask also a personal question. You're building one of one of your best friends, Alex. You guys worked together for 5, 6 years. You were super close. And then your wife is the COO.

30:40 You're building so closely with her. It feels like there's probably no line between personal and business at this point. You're working with your family and your best friends. What is that experience like? Like does it get difficult sometimes and challenging? Does it let you move faster since I think everyone has so much love and respect for each other at the company? It's so palpable. But what is that all-encompassing experience like running Serval right now? >> I think it lets you move just so much faster cuz it feels like there's infinite bandwidth. Like all knowledge is shared all the time and there's total alignment. You know, that you're not worried about getting people on board. you're not worried about convincing other people. You're you're kind of fully aligned. Not to say that there aren't disagreements and discussions, but there's total alignment in terms of what we're trying to do, what we're trying to accomplish. I think with my wife in particular, we were always the people we met at an entrepreneurship group. So we're always the people that are talking about work regardless. And before Serval, before she joined Serval, we're talking about our own worlds all the time, right? So I'll talk about what's going on my day, she's talking about what's going on her day. And it's kind of like somebody telling you about a TV show that they watch, but you don't watch. And you know, you you want to be interested, you're trying to follow it, but you don't actually know any of these characters and none of this really makes sense. And and so the conversations were just a little bit less meaningful because you're just trying to keep up with like what's going on. You don't watch the show. You're not there.

31:57 And now it's like we're talking about our favorite TV show because it literally we're living this all day long and so all the conversations are intrinsically interesting and all these things that I don't know that she saw during her day, that she doesn't know that I saw during my day. So I think it's been great for us, but I think it it is great because we started from the perspective of we're talking about work anyway. >> Right. I think there are some households that just don't really talk about work at home and I think this would definitely hurt that dynamic.

32:24 >> >> That makes a lot of sense. And maybe a last question around this point is what kind of keeps you guys up at night? Like what's the hardest part right now? A lot of applied AI companies might tell you that, you know, the token costs and the cost of compute obviously going up when a lot of folks were hoping that would go down and you know, it's there's a lot of new cost in the system running these models. Many many have moved to open source models and kind of fine-tuning those using their own data to address those costs. Is it a cost question? Is it a go to market question? What are sort of the biggest challenges you face today?

32:53 >> It's not a cost question because our model, you know, we're not really in the business of reselling tokens. You you use some of these when you're building out these workflows, but then these workflows get stored as tools that can be called again. So you're not kind of every request doesn't involve going and building out the code from scratch to go and automate that. You're referencing in most cases a tool that already exists. And so relative to other AI companies, I think our our token costs are quite low.

33:18 >> Totally. >> What I think about what keeps me up at night is that everything is moving so fast and this idea of like finding product market fit and then keeping it is just it doesn't exist anymore. The the market moves so fast and so the product has to really be sprinting alongside the market >> Yeah. >> and it's also evolving so quickly that you're kind of always going to be trapped chasing product market fit and that's not going to go away. And just accepting that we're going to be constantly chasing it and there's not a moment like maybe you had 10 years ago where you could build a product and have many years where that product would have like this this really tight fit with what the market was looking for. I think today this is shifting so fast.

33:57 >> Yeah. >> And that's you know, gives you this paranoia of like we got to we got to move fast enough to keep up with it. >> Well, it also plays into your strengths I think the team you've built, the fast feedback loop with customers, what you're doing with future founders and this forward deployed engineer motion. I think you guys are really well set up for how fast-paced it's moving and just on behalf of GC, it's been incredible to watch the last couple years and how far you've come since you know, brainstorming on paternity leave. So congrats on everything.

34:22 >> Thank you so much, Mark. It's been great to work together from the very >> Likewise.