From Tickets to Autonomy: Reinventing IT with AI Agents | The Pair Program Ep87

Speaker A: Welcome to the PEAR program from Hatchpad, the podcast that gives you a front row seat to candid conversations with tech leaders from the startup world. I'm your host Tim Winkler, the creator of Hatchpad.

Speaker B: And I'm your other host, Mike Gruen.

Speaker A: Join us each episode as we bring together two guests to dissect topics at the intersection of technology, startups and career growth. Welcome back to the pair program. I'm your host Tim Winkler, joined by my co host Mike Gruen. Mike, how was the uh, Thanksgiving holidays uh, for you?

Speaker B: Uh, pretty good, uh, seemed to be a little more traffic than usual this year, uh, traveling up the east coast, but uh, uh, otherwise nice uh, relaxing Thanksgiving.

Speaker A: Are you, are you a guy that does Black Friday shopping? Do you, do you actually go?

Speaker B: I don't think I've, I don't think I've ever gone shopping on Black Friday.

Speaker A: Good for you. Yeah, yeah, I made the mistake and I ventured out just because we had to go to Lowe's, get some uh, Christmas decorations for the house and uh, absolute zoo like people, you know, fighting over inflatable snowmen and extension cords. It was an absolute horror show. But um, yeah, I mean I come in, there's nothing, nothing you can do about going out in the Black Friday. Well, cool. Uh, enough about the uh, chaos of the holidays. Let's, let's talk a little bit about the chaos of the day to day grind of the IT service management industry. So we're going to be talking about a space where everything is uh, mission critical, somehow still painfully manual for a number of folks. Uh, tickets getting routed, rerouted, misrouted, um, automation kind of feels a bit half baked if you will. And uh, the tools that most teams rely on haven't evolved at the pace that the rest of the enterprise has. So uh, that is starting to shift and uh, once start startup, uh, Serval is right at the center of that shift. Serval uh, is building an AI native IT platform where agents don't just assist, they actually take action. They've got a code generated engine powering workflows, a purpose built modern ticketing system and I bet that autonomy, not dashboards, will define the next decade of it. And so we've got two great guests that can speak directly to this movement. Uh, first we've got Jake.

Speaker B: Jake.

Speaker A: I'm sorry, I should have asked. This is it. Stouch.

Speaker C: Stauch.

Speaker A: Stauch. Uh, Jake Stauch, CEO and co founder of Serval, the guy behind the big swing, making uh, IT systems intelligent, uh, and uh, joining him is Patrick Chase, uh, partner at Redpoint, an early investor and someone who's evaluated a lot of companies trying to disrupt this space. Uh, Jake and Pat, welcome to the Pair program.

Speaker C: Thank you so much. Great to be here.

Speaker D: Yeah, thanks so much for having us.

Speaker A: Awesome. All right, now before we kick it off, uh, we start with a quick segment called Pair me Up. Here's what we just go around the room, toss around a pairing of two things that go together. Uh, Mike, you always lead us off. What do you got for us?

Speaker B: So, uh, uh, ever since we were little kids, we always did a family photo at Thanksgiving. So family photo and Thanksgiving and it's all the kids all lined up. Um, and we've just been doing it year after year since I was a little kid. I appeared at my first Thanksgiving at I think three days old. And um, and then, uh, we've just. That tradition has kept on going and it's gone on to. So my parents do it and now my, my kids are in the picture and it's just nice. It's nice little family tradition.

Speaker A: We, we go with matching outfits. What's the dress attire?

Speaker B: No, you just, whatever you wear to Thanksgiving. Like, it's not.

Speaker A: Oh, I gotcha.

Speaker B: It's just on Thanksgiving Day we just take a family photo and it's just we have a nice collection going now for 50 over, I don't know, probably 60, 70 years. I don't know. However old the oldest of my cousins is.

Speaker A: You roll those into, into holiday cards or do they, they serve that purpose?

Speaker B: Uh, not usually, no. They're just a nice little. We just have them.

Speaker A: I honestly, I've never seen, Never seen.

Speaker C: We don't.

Speaker B: Well, that's because we don't send you Christmas cards.

Speaker A: Yeah, okay.

Speaker D: Just kidding.

Speaker B: Actually, no, uh, I think we probably do. And we probably send them out in January. We're, we're pretty late with our cards. We always, it's always a New Year's card, not a holiday, uh, card. So.

Speaker A: Nice.

Speaker D: Yeah, yeah, that, that sounds like an amazing tradition. And also complete chaos. Jake has a toddler. Even just getting one child to look at the camera is a lot of work.

Speaker B: So yeah, we're at a sweet spot now where the youngest person in the photo is 15 or 16. Uh, so our kids are all younger. But then. Right. It's only a matter of time. There was like, there was chaos. Then it leveled out. Chaos. Now it's leveled out. It'll be chaos again soon.

Speaker A: Pat. I've got a two year old daugh, a 100 pound dog and so we actually just did holiday pictures and it is like hurting cats, man. It's like trying to get the dog to look at the camera, you know, so somebody's doing like a treat thing over here and it's absolute, yeah. Chaos. But um, it worked out. All right, good stuff, Mike. All right, I'm gonna jump in. I'm gonna go with uh, Christmas lights and false hope. Um, as I mentioned earlier. Yeah, we were decorating the house. Uh, so I'm putting Christmas lights up on the house, feeling festive and I uh, plug everything in and uh, of course have a whole strand that's out. Like a true Clark Griswold moment where all that work was just completely humbled by a twenty dollar string of lights. So, um, I'm gonna go with Christmas, uh, lights and false hope and uh, yeah, pass it around. Um, Jake, quick intro and your pairing.

Speaker C: Yeah, so I'm Jake. I'm the co founder and CEO here at Serval. These are all great ones. We did do a, uh, Christmas, ah, family photo after Thanksgiving. We also have 100 pound dog and a 2 year old. Funny enough, it was challenging getting them to smile. And uh, I talked about the fact that I have never put up Christmas lights and, and we don't own a single decoration. We just instead, uh, go to other people's houses where they decorate. Because the thought of that, uh, yeah, I've abandoned hope. There's uh, nothing abandoned hope.

Speaker A: That's good.

Speaker C: So, uh, my pairing, I start every day and I finish every day and I have in the middle of every day basically a protein shake and a energy drink. Uh, side by side, uh, at my desk at all times. I've got, I'm always sipping a protein shake and always sipping an energy drink. And it's basically my like, unfortunately, probably 90% of my calories at this point are just like a combination of protein shakes and energy drinks. And it's kind of like alternating between the two. Yeah, you can subsist at least for uh, so far. You can subsist for quite a long time on, on protein and energy drinks alone.

Speaker A: I dig it, man. Yeah, I'm a big fan of, of both of those. Is there a Go to energy drink that, that gets you going?

Speaker C: There is actually. My wife started an energy drink company called Mati Energy. And so she uh, doesn't own anymore. She ended up selling it. But I, I am addicted because that's the only drinks we had in our house for many, many years while she was running that company. And it's still around. So we Order it now to the office. So our fridge is stocked with Monty energy drinks.

Speaker A: That is dangerous. Your wife running a energy drink company and you're being a startup founder.

Speaker C: We had it on tap at one point, so we had a kegerator in the fridge and this was coming out of college and it was the only drinks we had. And so it's, you know, it's 10pm, it's 11pm, it's midnight. It's the only drink we have in the house. So you're just kind of non stop chugging energy drinks.

Speaker A: So, yeah, the serval success makes sense, man. You're the most energized entrepreneur out there. That's hilarious. Uh, cool, man. Well, thanks again for joining us on the podcast. And Pat, yeah, quick intro and your pairing.

Speaker D: Yeah, I'm Patrick Chase. Uh, thanks so much for having me. I'm, uh, managing, uh, director at Redpoint. Focus, uh, on, um, infrastructure and AI applications and, uh, excited to be here. Um, the pairing that I was thinking of is also holiday themed. And, uh, I was thinking of Love, actually, and decorating Christmas, um, or, uh, doing holiday decorations. It's kind of been a tradition in our family that my wife started that we always put on Love, actually and then start doing all the holiday decorating. And so, um, that just always makes me think of, you know, family and that time together. And, uh, it's a great movie that is awesome to go back to every year.

Speaker A: Uh, such a good one. Is that, um, Hugh Grant? Is that.

Speaker D: It is Hugh Grant. He's, he's the prime minister. I believe in that one. And then, um, there's a really great cast of, I believe, mostly, uh, mostly English actors.

Speaker C: Nice.

Speaker A: M. Nice. Did you get a lot accomplished? Did you, did you knock it out or you still. Still a lot of work to do.

Speaker D: So this year we have a toddler that's helping, as I mentioned. So we have a lot of work left to do. Um, we also. This is a little bit down the rabbit hole, but we keep increasing the size of our tree each year. I'm trying to see how big of a tree we can fit into the house. Uh, and this year we got the biggest tree that we've ever gotten. What it meant was we've barely started decorating it, so we have a lot of work left to do.

Speaker A: I feel like you guys are gonna eventually switch up the movie to National Lampoon Christmas Vacation where someone pulls in the tree that's just completely.

Speaker D: Yeah, like the holiday is also high on our list. But, um, yeah, once. Once, uh, our Toddler gets more opinionated about what is on the tv, then we'll, I'm sure, be watching kids, uh, movie or something.

Speaker A: Yeah, the switch to the Grinch. Good stuff, man. Well, well, thanks again for joining. Uh, and uh, that'll wrap up the uh, the Pair Me up segment. Let's dive into the, the heart of uh, today's discussion. So we've got three areas that we really want to kind of hone in on. Uh, first is kind of like the core it uh, bottlenecks that Serval is setting out to fix. Second, um, I want to dive a little bit deeper into some of the AI native architecture, uh, behind the platform. And then third, yeah, how this all kind of stacks up against the legacy players and IT service management. Um, so let's kind of set the stage here. Jake, I'll start with you. Um, you know, you saw firsthand how messy and manual, you know, IT service management still is. Uh, you know, what was the core pain point that you were really trying to solve when you started Serval?

Speaker C: Yeah, for sure. So I was working at a company called Verkada that made enterprise physical security systems. And I was director of product over there. My job was to talk to IT leaders and ask them about their problems and figure out what we at Verkada could go build and solve for them. With a focus obviously on enterprise security. So building cameras and access control systems, alarm systems, and all those conversations would start off with some version of hey, what can we build for you next? And the IT leader responding. Well, you could get me out of the help desk for one. We just heard that over and over and over again of, uh, get me out of the help desk, you know, automate my tickets, you know, automate my onboarding, my off boarding. And one conversation in particular really stands out where he, this, this IT leader explained to me what this process was for fixing an offline camera. She's like, here's what happens when a camera goes offline. Um, somebody reports it. That becomes a ticket and servicenow or another ticking system and then that gets assigned to somebody. Eventually that person investigates a little bit. They'll look up the camera, they'll, they'll look at some, some stats, then they'll assign to somebody else. Eventually somebody will look at it and you know what happens is two weeks later somebody reboots the camera and the camera's fixed. He's like, why does that take so long? And can you guys just reboot the camera automatically whenever a ticket shows up in ServiceNow? And um, that was the that was kind of the nugget of the idea of, oh, it would be pretty cool if you could have some kind of system that identified these, these issues as they're coming up and just rebooted the camera, rebooted the other device or called a technician. And so what the initial, uh, concept of Serval really started as we could have this layer that sits on top of ticking systems, identifies initially network device issues. That was our focus, since we were so close to that at Verkada, and then resolves those. And eventually as we talk to more and more customers, we realize the scope of this problem was so much bigger than cameras, than network devices, that you could actually automate resolution of a lot of issues. Um, but that was the beginning.

Speaker A: That's cool. Yeah, I was doing some research, um, on your background and listening to some other episodes or some other podcasts that you were on and some other articles that you were featuring in. And there was, um, ah, a line there that you had mentioned, uh, to pull people out of a legacy system. Um, uh, what you build has to be 10 times better, not slightly better, you know, So I was wanting to pull on that though. What does 10 times better actually mean in the world of IT service management?

Speaker C: Yeah. When you're going after a legacy category, you have to think about why would someone possibly want to rip and replace a system? It's not easy to do that. Right. And so you have to attack it from an angle that just provides so much exceptional value. And so what are you trying to do in ITSM M that's super valuable. Yes. There's ticket tracking. Can you build a better way of tracking tickets? Maybe, uh, you could certainly build a better ui, but that's not really going to get people to rip out their system. The other thing people are trying to do with these systems is automate certain workflows. They want to be able to automate the assignment of tickets, maybe automate some of the troubleshooting steps, um, some of the rules around, follow up. So they build all these workflows, these automations, to actually take action on the problems and requests that are coming in. And so if you want to replace those systems, what you really need to do is build a better way automate those requests. And so we decided to really focus on automation and not so much build a new ITSM as build a new platform for automation and then build the ITSM around that. Automation was really the problem we're solving. And our idea around that was, yes, there are a lot of tools to build automations, but the pain point is there are too many tools to build automations, they take too long and they, they take too much of a time investment and a uh, in an ROI calculation you have to make on whether or not it's even worth it. And so what we decided to build was a way to automate the automation to make it so that you could describe what you're trying to automate. And that automation, that workflow just builds itself and that way we take the friction out of it and that allows you to actually start to automate these help desk requests at, at scale. And so the way we decided to tackle ITSM and this legacy category was going after a piece of that that we knew would unlock a lot of value.

Speaker A: Yeah, it's very interesting Pat, I wanted to get your perspective on this. As somebody who's probably evaluated a ton of itsm um, startups over the last few years, what was it that stood out uh, about Serval, the moment that you kind of met Jake and kind of saw the early product?

Speaker D: Yeah, for sure. I mean it was um, we've been looking at the ITSM market for over two years. Um, we kind of were really excited about a next generation solution uh, in the market and we can talk about the reasons why. But I think the few things that really stood out to us about Serval I think first is obviously the team. Um, you know Jake and his background and um, kind of deep domain expertise coming from Verkada and having seen this pain point first and foremost, um, was something that, that we were really excited about, um, the technical depth of the team, um, you know Jake's co founder Alex and all the folks that uh, that came over with him. Um, and then another thing we really look for in early stage teams is, is talent density and the ability to recruit from uh, you know, recruit talent from places that these, these people have worked uh, at in the past. And that was kind of off the charts with Serval with a lot of people coming from um, Verkada and places that Jake has worked and Alex and now Rippling and the rest of the team. And so um, I think domain expertise plus talent, uh, density and gravity were two things that uh, really stood out to us. Um, and then there was a lot uh, on the product side that was aligned with kind of how we see the world. I think the first thing there was Serval was taking this full stack approach of being system of record in this world. That's the ticketing system plus system of intelligence and building both of them together. And uh, we really think that's what it takes to build a large standalone company in the space. Um, and Servol was doing that from day one, although they didn't make customers adopt both. But the fact that they did have both and had the long term vision there to be full stack was really important to us. And uh, then the last thing was that they were code native. Um, they kind of were built from uh, from day one to have all these workflows live as code. And um, instead of having a human configure everything, it would be uh, an LLM generating code. And um, there's a lot of advantages to that as well.

Speaker A: Very cool. Yeah. Jake, you know, with the goal of automating, you know, a task faster than a human can do it manually, you know, how did you. Yeah, how did you kind of architect Servol to realistically make that possible?

Speaker C: Yeah, for sure. One of the conversations we had when we were doing customer discovery was really interesting. It was a guy who had built out this crazy workflow and okta workflows and he described to us what he was trying to automate. He said, what I'm trying to automate is when someone submits an expense, I want a manager who is an M M5 manager. It's a certain ranking of uh, manager. An M5 manager or above has to approve that. But if that means going all the way up to the CEO, you've gone too far and you got to drop down a level, even if that's an M4 manager. So he said that in one sentence and then he showed me the workflow he built to do that. And he had to scroll maybe for 20 seconds through pages and pages of this workflow diagram built uh, out in okta workflows, if this, then that rules engines, uh, error handling. And he showed this to me and he's very proud of what he built. He built it over a period of several months. And I just remember looking at that and thinking, pat, there has to be a better way than that for something so simple. And what struck me at this moment and several moments since is that you can almost always describe these workflows in one sentence and then you go to build them and it ends up being all this crazy logic because you have to handle all these cases and uh, all these rules. And so what we started with from the beginning was what if you could just do the sentence. So you could start with that sentence of uh, here's what I want to do and then use AI to build out all the underlying logic and error handling and all the things that have to go into making that a robust workflow. Now the only way for that to really work is for you to let the LLM do what it does really, really well, which is write code. And so from day one we decided to take a huge bet that code generation was going to be good enough and was going to continue to get better and better so that you could describe what you're trying to automate and then have that build out as a, uh, code based script and have that be reliable and really easy for people to implement. And uh, part of that was just a bet that we took that the technology was going to get there because it certainly didn't work day one. But we kept investing in it and the models kept getting better and eventually we got to a place where now it's incredibly reliable. And you can just describe that workflow that that guy described and it'll build out that full end to end system in serval, um, in a matter of seconds.

Speaker A: That's cool. Yeah, I want to dive deeper into the code native side of things. I think that's really fascinating. But uh, while we're still kind of on that topic and the early days of building, were there things that you kind of threw out because it was slowing you down? Were there things that you maybe thought was a good idea in the early days and then. Yeah, later throughout.

Speaker C: Yeah, I mean there's so much of that. I mean one of the biggest things that we had to figure out is, um, we thought it was really cool to have these natural language steps and then the code steps and we wanted them to be one to one, linked really nicely so that each step in your workflow related to the code step. And you could kind of like click into the step and it would show you the code and then you could like modify the step and the code would like automatically update itself. And that ended up just being so tricky to get right and to get at the right level of granularity. Because what happens if you have like a loop, uh, or you've got some kind of like if this, then that, like how much of that do you represent in the natural language steps versus how much do you like abstract away? And we just, we got ourselves so caught up in this debate that we ended up throwing that away and we said, you know what, we'll take your natural language prompt that will generate the code and then we'll show you the natural language steps. But it's all made up and derived from the code. So the natural language steps are basically us Taking an LLM and saying describe what this code is doing step by step versus having there be necessarily like this one to one relationship where every line on the page is actually is a representation of some piece of code. Now it ends up more or less aligning in the same way. But uh, it works a lot better to just kind of not have that constraint of every single line has to map perfectly to what the code is doing. Instead you say, hey, describe what this code is doing in easy to understand language and easy to understand steps, and then represent that to the user. And that ends up working out a lot better. Um, having just that m. That like middle piece be purely a visual representation and not actually something that click into and configure.

Speaker A: Yeah, it's interesting, Mike, I'll pass to you. I know that.

Speaker B: Yeah, I mean that's fascinating. I think it, uh, totally makes sense, right? Like just that in of itself. The, um. Because I imagine it also gives you feedback loops in terms of like I, when I natural. I used to do natural language processing for trying to find bad guys. Uh, and uh, you know, being able to take that it would generate a bunch of code and generate a bunch of logic or whatever and then translating it back, being able to see where the holes were. Right? Because now you're also getting that feedback loop of describe what the code is actually doing. It's like, oh, wait, I know I said this, but it looks like it was interpreted as that. So I could see that. Um, I'm curious like from a. So this is for it from like a security perspective and other things from an auditing perspective. I imagine there's a lot of advantages to this as well. And I'm curious if you could maybe talk about that.

Speaker C: Yeah, for sure. If you think about traditional workflow builders, they often become a black box. You don't really know what's happening in the middle there. Uh, the workflow application is generally abstracting away a lot of the logic. So you have these blocks, right, with certain inputs, certain actions, certain outputs, and they're these standardized blocks and you can kind of mix and match blocks. But what you end up with is just kind of like uh, an abstracted new programming language that oftentimes is just as complex as writing underlying code, but takes up a lot more space because you've abstracted it into this like no code interface. And then you don't really know what's happening in those blocks. It's kind of, it's completely abstracted away from you. Um, so from an outside in perspective, there's a lot of question marks of like, okay, what is that workflow actually doing? Because we've taken this code based approach and because of a couple other things we've done including building our own um, durable execution system. We have inputs and outputs on every single API call that's made. And so you can actually follow a workflow and see every single step in the workflow, what API call uh, was made to which application, what are the inputs that call were, what the outputs were, and all of that is logged. One that's great for security and troubleshooting and transparency that you know exactly which code, what code is being executed, you know exactly what's going into and out of all the different application API endpoints. But then furthermore, for troubleshooting and maintenance, you can see, okay, why did that go wrong? You actually see exactly, oh, you couldn't create an email alias because you got an error back. That said, you've created too many email aliases versus the way these, these traditional workflow builders. It often silently fails. You go to run the workflow, nothing happens. Uh, you don't really know why that's

Speaker B: that or, or you get an error, but it's totally meaningless and has nothing to do with what the actual underlying problem is. I've never, never experienced that. Um, um, and I'm kind of curious like so you mentioned legacy and going after legacy markets. Is there because. So I've worked almost entirely at startups most of my career. I've also always been, in addition to being the VP of engineering, responsible for like it and whatever it usually rolls up into me. Um, so being in, it's easy as a startup, um, to adopt these new systems and roll these things out. And like you mentioned ricotta, that was one that we used. Um, Pat mentioned rippling use that. Um, you basically described one of the stacks. Um, anyway, so um, and as a new company, like I remember we'd have employees join from these big enterprises and they'd be like, oh my God, you're it. It's like you guys are so modern and cool, like everything's automated but it's so easy when you're new and starting from scratch. So are you seeing a difference in like the adoption for new companies versus legacy? Are you primarily going after those legacy customers? I' of curious like how you, how you sort of look at that space.

Speaker C: Yeah, and the, the, the real market is a legacy customer. So certainly we have a lot of customers that are new, fast growing AI startups. They move very quickly so they end up being the first Customers because the single cycles are not very long. Um, right.

Speaker B: Because something is better than nothing. You're not trying to, you're not, they're, they're introducing a new thing, right?

Speaker C: In some cases where their very first ticketing system. In other cases they've implemented a solution, but they only implemented it a year ago. And so it's, it's much faster to go and replace it with something better. M and then their business processes, their procurement, uh, everything just moves very quickly. So that ends up just by default, by pure speed, being the early customers. Um, but our intention and our focus is really on the large enterprise. If you look at um, ServiceNow, which is obviously the largest player in the space, most of their revenue, I think it's 60% comes from contract sizes, over a million dollars in uh, annual contract value. So the market uh, for this category is all in large enterprises which are traditionally on legacy systems. So that has to be the focus because there's just not enough in the, in the mid market SMB space to really build the kind of generational company we're trying to build and actually to go after ServiceNow which is the leader. So that's, that's our focus now. That ends up being challenging in a few ways. So one, as you mentioned, uh, the tech stacks are a little bit older. Uh, that ends up being okay for us because even the older tech stacks generally have an API. So workday API is not as nice as the rippling API. Um, but it still can do all the things that you need to do. So you can integrate with that and you can automate a lot of cool things. Microsoft APIs is uh, pretty robust. And so because the way our workflows are built, as long as the APIs exist, you can build any automation you can imagine. You can automate any action in any application as long as you have an API. And so the legacy systems don't matter as much. What ends up being the biggest point of friction, um, besides procurement and deal cycle and all that, is actually that the information on how things work today and how they should work is distributed amongst a much larger set of people. So if you're talking to like a high growth tech company, you can generally find one person that knows exactly how things work and exactly how they should work and should be changed to work. Uh, when you're talking to a larger enterprise of maybe 150,000 employees, that information lives in many, many people's heads, some of which don't even work at the company anymore, some of which people don't know how to find. And so it ends up being more of a game of hey, what do you want to happen? That ends up being harder than actually building it so that it works.

Speaker B: M and um, you mentioned um, yeah the sort of if there's an API you can automate it. Are you guys looking at doing any other sort of operator? Like if there's just a UI we can plug into that where there is no API? Because there's plenty of legacy systems in it where there's. There's no API.

Speaker C: Exactly. That is a big focus of ours and something that's on the roadmap is the computer usagents, the browser automations. Um, I think it's a super interesting category and something we definitely need to do because as you said there's a lot of things that don't have an API. The challenge there is really around security and determinism. And so one, how does the AUTH model work? Uh, you give this system username and password into a super admin account that can do all these things. That's a little scary. Um, you know how do you handle multi factor authentication? And that's something you want to do that, that ends up causing a lot of heartache. Uh because then somebody still has to babysit the process which uh, kind of takes away a lot of the value in the automation. Um, and then the other thing is the determinism versus non determinism and this is actually the much harder problem to solve which is okay as long as nothing changes about these applications. You can go and you can automate these things, you can have a browser interact with these applications. What happens when you change that menu and you move this button to the left and you change how this button works? One, you could just say that the automation breaks and doesn't work anymore. That's pretty frustrating because it means that your automations are only as robust as long as no one changes their website. Um, but two is the alternative is let the LLM decide how it should handle this case. And that's kind of scary because uh, maybe you had a uh, Create user button, uh, that used to just create the user but not assign a license. But maybe the application changed it. So now Create user also creates a license which might you know, cost the company money. There's all these cases where you let non determinism creep in and you expose yourself to all this risk.

Speaker B: Yeah, I can imagine. I mean on those changes uh, sometimes people just copy and paste UI code and so that menu, the, the met M on the back end, it still Looks the same. Like if you're just looking at the HTML or just the, the form, it looks the same. But like if you were a human you would see like actually this is, this has completely. This is not, this is not the user that is over on this other thing. I've run into that problem plenty of times. So uh, yeah.

Speaker C: And maybe now the Create user has become a. Delete all users.

Speaker B: Exactly. Right, right, right. It says create but. Right. The name of the form is Create user, but it's not doing that anymore.

Speaker A: I want to uh, jump in again from like the investor, uh, lens because a lot of the things that you know, Jake and the team are doing is pretty unique in the space. Like the like not just being like a pure AI overlay. Right. They're um, like the foundational side of owning both the system of record and like the system of intelligence. Like that whole full stack approach is, is really interesting. But also the code native piece. Were there things that jumped out? Um, I guess when you were. Yeah, you know, uh, evaluating the, the opportunity to, to collaborate with these guys, you know, what were the things that you were kind of like putting on the priority sheet that said like this is, this is the category now that can really be disruptible.

Speaker D: Yeah, I think um, taking a step back, I mentioned that we've been looking at the category for two years. Um, and the reason why was we felt very strongly that there would be a next gen IT service management player, um, and they would likely ride this um, wave of code generation. Uh, so why do we think that? When we looked at ServiceNow, there were kind of two main moats that we saw. The first was customization and the second on integration. And when we looked at customization, obviously they're incredibly powerful product. Um, Jake was talking about how many huge contracts they have. I think they have um, over 2,000 customers paying them over 1 million ACV and over 500 customers paying them over 5 million ACV. Um, it is really, really impressive um, what they've been able to capture at that part of the market. Obviously there's a ton of complexity that comes with that and their um, ability to bend the software that they built to fit any enterprise super powerful. The way that they did that was allowing humans to configure everything. And I think the next gen version of that will be AI configuring everything with, with co generation and these applications being built on the fly. So we were looking at, okay, if customization is a, is a really strong moat, maybe there'd be an opportunity for a Next gen player to take advantage of that with code generation. Um, then we go to the second moat around integration and again Jake was talking about, you know, how you can write integrations with code to workday, API and those types of things. Uh, and so I think there's also an opportunity for a next gen player to, to ride that wave. Um, and so I think those two things aligning, um, you know, really create a, you know, a generational opportunity in this market and that's what we saw and we feel really, you know, fortunate to be part of the Serval journey going after it.

Speaker B: Yeah.

Speaker A: And how much of that is timing too? Like um, the LLM maturity, um, kind of lining up right now as well. Talk to me a little bit about the timing of uh.

Speaker D: Yeah, I think there's a lot of it that's timing. Right. I think if you were building an IT automation platform, there's been many in the past that and your time, you know, you had found it a few years before, um, Jake and Alex founded Serval, you probably would have architected it in a different way. Um, it may have been a lot more NLP based, kind of traditional NLP type, um, parsing and processing of these tickets versus just throwing everything at an LLM. Ah, so it is hugely timing based and I think that is these models have gotten so powerful and there's a reason why it's all starting to work right now. Um, Jake was touching on one of the problems that they tackled in the early days of having this one to one mapping and then how do you summarize it to make it human readable? It's an interesting case of something we've seen a lot across the portfolio which is that problems, the answer to problems with AI is often more AI, um, where you know, if the AI is not working then you add an AI model that is overseeing it or you add an AI eval. Um, and so you know the, the models are, are really, really powerful now. And I, I don't think you would be able to build a Serval two years ago or three years ago. I don't know. What do you think Jake, do you agree with that or disagree?

Speaker C: I don't think you could build what we have today, even six months ago. I mean I think a lot of the stuff we do today, uh, is a lot less reliable six months ago. And we could not be winning the accounts and having the conversations we're having if the models had not continued to evolve. And so there's a little bit of a bet that has to happen when you're, when you're building this kind of space that you're kind of building at the edge of what the models are capable of and not deluding yourself that, oh, everything's going to be fixed by ChatGPT 5.2, um, but also kind of pushing the boundaries of what's possible in the current generation of tools.

Speaker D: Investors can believe that everything will be fixed by ChatGPT 5.2, but, you know, founders, they have to actually solve the problem. Um, so, yeah, yeah, the timing thing

Speaker A: is really fascinating because, you know, we talk about a lot of AI in, you know, critical infrastructure verticals, if it's, you know, for war fighters or, you know, health care, energy. And just at the, the pace that, uh, the AI is just accelerating. It's wild, the bets that a lot of these founders have to make with the anticipation of what's about to happen. But it's changing so quickly. I wanted to ask if there was any parallels that stood out to other markets, maybe where redpoint has backed some challengers that are going up against massive incumbents like a ServiceNow.

Speaker D: Yeah, I think it's really deep in the Red Point, kind of our philosophy, um, of taking big swings and going after large markets. Uh, we were fortunate to, um, lead the Series A in Snowflake. And I think there's a lot of parallels. Um, obviously the platform shift, there was cloud and then going after, um, the cloud data warehouse market. Ah, but I think, you know, huge enterprise budget and market and opportunity, very, um, entrenched incumbents. And um, you know, as Jake was talking about, you have to build something that's 10x better to go win those markets. And um, you know, I think it, it can take a lot of capital and time and um, you know, can be a challenging road, but then the prize at the end is really big. Uh, and so we, we like to partner with those types of founders that have that vision and ambition. Um, and you know, we know that amazing, uh, companies take time to build. Although Serval's been building everything really quickly and moving, moving up market a lot faster than we expected. So that's always great. Um, but I think it is, uh, yeah, it's similar to a lot of the, you know, kind of big, um, swings that we've taken in the past.

Speaker A: Jake, I wanted to, um, uh, dig a little bit deeper into, you know, how customers maybe interact with, um, AI agents, um, you know, anything that surprised you, like, were there tasks that customers, you know, trusted agents faster than you expected and then where are, where are they still Kind of hesitant to trust the agents.

Speaker C: Yeah, that's a, It's a great question. Because in the early days we were so concerned that no one would trust that the workflows would work and they would need all this proof that the workflows would work. And instead they just did. They just built the workflows and they were shocked if they didn't work. And they just instantly published them and there was very little testing. And so I think that's been a pleasant surprise, has actually helped us move very quickly in that generally when the system says this workflow is ready to go and it's done, our customers just trust that it's ready to go and are comfortable running it. So the trust gap has been less than we assumed it would be. Um, our customers, because the platform is so flexible, they end up building a lot of things that surprise us and things, uh, that we didn't think were possible. And that's happened now just dozens of times where they show us some workflow that they built that we would have sworn is not possible based on how we understand the platform to work. Uh, and we often have, like, in the early days at least, we were like, no, that's, you misunderstood. Like, it's not working the way you think it is. And the customer literally have to share their screen and show us that they did build the thing that they said they built. Ah, one cool example was, uh, we, you know, they were able to reset or provide a customer, uh, an end user, their file vault recovery key to get back into their MacBook after they'd been locked out. Um, which, you know, totally understand that our, we have an integration with an mdm, then we can like pull that information, we can send all that. So that made sense to us. But they're like, well, we didn't trust that they are who they say they are. So we made them first, uh, get a push notification to their phone and then only when they confirmed on their phone that they got a push notification, then, then we'd send them the follow our recovery key. We're like, yeah, I don't think that that's going to work the way you thought it would. And then it just did. And all they had to do is say, hey, send an octopush notification to their phone and make sure that they confirm that before you send the follower recovery key. And our system just handled all that logic and it just made it work because the APIs for those applications support it. Um, so that was really cool. And the other one was, uh, this customer wanted predictive analytics on computers that were going to, that were struggling with Zoom calls for whatever reason. And so they're like, yeah, we want to, we want to just like pull a report on, you know, computers with high CPU utilization during Zoom calls and proactively like reach out to people, uh, with, uh, with some messaging around what they can do to improve the Zoom call experience. And we're like, okay, I'm not really sure. And then again they just showed it to us and like, oh yeah, it, it found the Zoom API, like outputted a report. It like even broke down the, like the list of users. It even like inserted emojis on like the health of their Zoom calls. And then you could like, LLMs are

Speaker D: very good at emojis.

Speaker C: And it was like, he's like, yeah, now I have this report that sent to me like weekly and automatically reaches out to people and I just had to like it. One shot at a prompt into Serval to set that up for me.

Speaker B: Does it also just automatically close all their Chrome tabs? Like you have 500 open. That's, that's the problem.

Speaker C: You have a customer that uh, when they complain about their computer, it sends a reboot command to their computer. So, uh, we have automated the turn it on again and turn it off again. It uh, support request.

Speaker B: Yeah, no, I'm sure you have. Uh, my favorite was new, uh, employees that uh, would lock themselves out of their laptop like three times in the first two days. It's like, I don't understand how, how you did that.

Speaker C: Really, really complicated password manager.

Speaker D: The one thing that, that was really cool, uh, about the server platform when, when we saw it, was the fact that, um, you know, not only is all of this code generated, but it's all exposed to the user, so user can go look at it and see every single line of code that Serval's written. And I think that goes a long way towards, um, building trust, especially with uh, a more technical buyer.

Speaker A: That's cool.

Speaker B: Yeah, I was going to ask about that a little bit in terms of like, feedback loops and other things. Like you mentioned the whole edge cases. Like it's easy. Like that's where engineering always is. Right? It's easy to describe the happy path. Right. This is how I want it to work. Is Servo also helping with the like, discovery of like, well, what about this case? What about this case? Or any of those types of things to sort of help build that trust that, yeah, we got the happy path, but here's all the edge cases you didn't think about.

Speaker C: Yeah, it's surprisingly good at figuring that stuff out. And also just representing a lot of this logic in code. You get a lot for free in terms of the error handling and all of that. And so a lot of that stuff just happens out of the box when you write these as, uh, scripts. The other thing that's really helpful is when you run it in Serval, say you build out, like, a password reset workflow, and then you run it and it goes and it tries to reset the password and it runs into some error. We get the error back, Serval gets that error back, and it automatically just says, oh, I see what the problem is. Like, let me go and fix this. Um, and it also has access to search the Internet. So it'll go and it'll fix that using the API spec, but then it'll search the Internet for help, you know, forums and API documentation and other sources. And so you can just kind of, even for a very complex workflow with all these different steps, you can just iterate with it, uh, and it will fix itself and it will solve these problems. And, you know, you get to, like, a really cool place at the end.

Speaker B: That's pretty cool. And I imagine, like you were saying, the, um, you know, the complexity, you know, these systems, you know, the things that have been there for a while, like, Gall's Law applies, right? Like, nothing. You can't just go in and build a complex system from scratch. You have to evolve. These things evolved over time. And are you guys doing things where you're pulling in a lot of that documentation and using that to generate the workflows and stuff like that? Um, I know you talked about it at the top with interviewing people, but I'm curious, like, at a system level.

Speaker C: Yeah, it's really interesting and we've thought a lot about that. What we find, though, is a lot of times the companies bringing in these tools, they admit that the processes are very broken. And yes, you could import some of these workflows, and we have some customers that do versions of that. But often what you want to do is reimagine how these workflows should work. And it's kind of the opportunity to restart this.

Speaker B: That's cool.

Speaker C: What's cool about ITSM is a system of record that's pretty unique as a system of record versus a, uh, CRM or an hris. Think about CRM, H, R I S and erp. You need all the historical data at your fingertips at all times, and it's all relevant. Like your pay stub for five years ago. You need it, um, the customer notes from the deal you closed, uh, six months ago. You need it front and center. Um, the password reset request from yesterday. Nobody cares, uh, it doesn't matter anymore. And so yes, you want all this data logged, but the kind of useful half life of IT service management data is so much uh, less than all these other systems record. And so you have a little bit more willingness to say, you know what, let's start fresh, we'll download everything, we'll archive all the old data, but let's actually start fresh from day zero on how we want these systems to work. And that seems to be a uh, much better approach than trying to map everything onto what are sometimes ancient processes.

Speaker B: That's cool.

Speaker C: I'll give you one example that I thought was really interesting is an onboarding process that this customer is looking to add. And they sent us this giant flow diagram of how onboarding worked. And we noticed that there are like all these crazy branching nodes. And when we, we looked at it, those nodes were creating new tickets for somebody to confirm the data from the previous step. So the customer like would literally send uh, in a uh, request, uh, to change their address on workday. And somebody would have a ticket that gets created that's like confirmed that user exists in workday. And then like so that's like a whole branching tree structure that then goes to somebody else and somebody approves it and goes down the structure. It's like, wait a minute, you can just look them up before the request goes through. And now that whole branch of that tree is gone. And then you, you, you go through each of these, you're like, wait a minute, this request is actually one API call. And you wanted us to build this entire workflow when really it's just like one API call into workday. And there's a lot of those processes that exist in ServiceNow today that are just gonna be completely replaced by just simple actions.

Speaker A: Mhm. I wanted to quickly ask, uh, maybe to you Pat, about like buyer signals, um, for, for folks ready to consider something new in itsm. Is this being uh, driven like top down, uh, by like CIOs, or is it, or is it bubbling up from uh, like the IT teams?

Speaker D: Um, I think the real answer is probably both. Um, but I do think there are a lot of top down. If you made me pick one, I would say top down. Um, because every leader, especially cio, is looking at their organization and their stack and looking for ways to leverage AI to become more efficient. Um, so I think there is a Lot of um, top down um, mandate to go improve the stack and improve the tools. And one of the things that we've seen in AI is companies are open to adopting these tools faster than we've ever seen in the past. Um, and um, buying cycles are even faster than we've seen in the past. And a big reason for that is the mandate that companies um, have from leadership and then just the value that these tools can bring. Um, and so it's kind of crazy how fast it's moving. You're asking about comparisons to previous platform shifts and those types of things. Uh, AI is a lot faster um, than the move to the cloud or the adoption of cloud tools and I think that creates a big opportunity for um, startups or the next gen version of them.

Speaker A: Very cool. Yeah. Before we uh, wrap with our, our final segment, I'll just kind of close on a ah, note uh, for you Jake. So kind of like looking ahead you know, crystal ball, three, three to five years out, you know, what, what does the itsm ah ecosystem kind of look like?

Speaker C: Wow, it's really hard to forecast that far because a year and a half cerval didn't exist. Right. So um, that's uh, that's an impossible time frame. And three, three years ago ChatGPT was launched for the first time. And so uh, yeah, if we look it out though, I think these shifts that we're seeing today will continue to accelerate where people are not going to stop wanting to customize these enterprise platforms to meet their business processes. And so I think the, the tools that are going to win in ITSM are going to be the ones that embrace that customization and actually make the customization and implementation of those customizations and automations very, very easy. And so I think you're going to see an explosion of automation um, because it's becoming so easy to build those automations and then you're going to see people take things a step further. And I think a lot of the what will start as simple help desk automations, you know, reset passwords, tell people what the WI fi is, add people to a Google group. Uh, we'll start to evolve more into, more into things that look like uh, actually doing work for people. So um, some IT requests turn into small IT projects and you can imagine those IT projects also start to get automated end to end by Serval. And so Serval goes from you know, workflow automation that resolves requests to workflow automation that handles onboardings and off boardings, uh, and maybe like sets up new rules in okta for assignment to groups and eventually kind of evolves into full applications or mini applications, um, that you use to manage your IT infrastructure. And so I think there's going to be an evolution where the tools that get built in serval are going to be increasingly complex while also being much, much easier to build and configure than historical uh, precedents.

Speaker A: Cool. Yeah. Well, excited to follow along and uh, we'll revisit this in three years and see, yeah, see how accurate we were with the uh, prediction. Uh, but I do want to make sure that we, we save a little bit of time for the five, uh, second scramble. So this is a, a fun little rapid fire Q A segment. Uh, we're each gonna do a, uh, you know, 10 questions, quick hitters. Try to give us your best uh, gut answer. First thing that comes to mind, some business own, be, you know, kind of fun personal questions. Mike, why don't you kind of lead with um, with Pat and then I'll, I'll close with, with Jake.

Speaker B: Sounds good. You ready, Pat?

Speaker D: Let's do it.

Speaker B: All right. If Red Point were an animal, what animal would it be?

Speaker D: Mmm, Husky.

Speaker B: Nice.

Speaker D: The Red Point is actually, it's a climbing, uh, it's a climbing term. Um, it means to kind of do a route after practicing it before. And Jake was actually one of the only founders that knew that off the bat because he is a climber and has a climbing background. But um, the reason I went with Husky is because one, mountains climbing, all of that. But two, um, they are very collaborative. They're always working together to pull the team or to pull the sled. And we kind of have collaboration and team orientation built into the structure of the firm and how we operate every day.

Speaker B: So Husky, very thoughtful answer. And now I'm going to have to ask my wife who's a, uh, climber, if she knows what that is. Um, uh, what is a, uh, common mistake companies make when seeking investment

Speaker D: optimizing for valuation or brand versus partner or who they'll be working with?

Speaker B: What's one impact AI is having that's maybe subtle or that others may find surprising?

Speaker D: I think the long tail of use cases, uh, how good it is for all types of random things, the co generation support. All of that is in the news every day. Uh, I think the part that's missing is just how great it is for planning a workout, helping with a parenting question, um, logistics, all of that.

Speaker B: Uh, what's one trait you look for in a leadership team before investing in them?

Speaker D: Ambition.

Speaker B: What advice have you received that's really stood the test of time.

Speaker D: I've been fortunate to receive a lot of good advice. Um, I think just being grateful. I, um, think we're really fortunate, uh, to be where we are in the world and to be in the center of all of this AI opportunity and AI value creation. And I do whatever I can to not take it for granted.

Speaker B: That's a great answer. Uh, what was your first job?

Speaker D: Uh, my first job was as a software engineer. I worked at, uh, at LinkedIn as an intern on their machine learning team.

Speaker C: Cool.

Speaker B: Uh, in high school, what was your dream car?

Speaker D: I wasn't really a car person. I was a. I grew up surfing a lot. And I've always really liked the, uh, Toyota FJ or whatever that was because you could hose off the entire. The entire. From the front. From the front, seats back. You could just hose it all down. And I thought that was really cool.

Speaker B: That's pretty cool. Um, what's your, uh, go to midnight snack?

Speaker D: M M's nice.

Speaker A: Peanuts.

Speaker B: Yeah, I just can't say plain or peanut.

Speaker D: Peanuts are my favorite. We often have different kinds of chocolate, so, um, it's one of those things where I. I never want to buy them, but then I'm always upset we don't have them.

Speaker B: Uh, if you had to teach a master class on something that has nothing to do with your job, what would it be?

Speaker D: Handstands. Uh, I got super into handstands and bodyweight training during COVID I wanted something that I could do just in our apartment. Um, and that had an infinite learning curve and I could never master. And definitely gymnastics strength training was that. Um, but I spent a lot of time doing handstands and falling, um, on my head.

Speaker B: That's a great answer. Uh, and last one. Uh, what's a charity or cause that is near and dear to you?

Speaker D: Uh, for me is the, um, the Trevor Project. Um, and yeah, I. We can decide whether or not to keep this, but, um, I guess Jake probably knows this or, you know, work together for a long time. But, um, my, uh, my sister was gay and she, um, she passed away in a car accident, um, a few years ago. Um, she was unfortunately, ah, killed by a drunk driver. And, um, you know, since then, we've always been big supporters of the Trevor Project. They, um, help with suicide prevention for, uh, the LGBTQ community. Um, and it's always been very important to me.

Speaker B: That's great.

Speaker A: Yeah, thanks for sharing that. We'll also promote that in the, uh, show notes when we push the episode. Cool. Pat, you're uh, good. Jake, you ready?

Speaker C: Let's go.

Speaker A: All right, uh, if Serval were a band, what genre would it play?

Speaker C: I was hoping for the animal questions. Animal.

Speaker A: Um, yeah, we actually talked about that. I got a pivot.

Speaker D: That would have been easy.

Speaker A: Yeah.

Speaker C: What genre would it play? That's a good one. I, I, I, I blast Taylor Swift all day long and so it's hard to imagine uh, Serval playing anything but Taylor Swift cover band.

Speaker A: Yeah, Serval Swifties.

Speaker C: Yeah, for sure.

Speaker D: I mean sir, it's hard to imagine anyone's reach or ambition being bigger than Taylor Swift.

Speaker A: So I think spot on. Uh, what's one trait that you screen for every single time that you hire at Serval?

Speaker C: Um, do I want to spend my day with this person? Because we spend a lot of time together and uh, even though it, it seems like uh, maybe unrelated to work, it's just like hey, this has to be somebody that we all want to spend our days with.

Speaker A: Yeah. To keep it simple and it's a good answer. What kind of uh, tech roles is Servo hiring for over the next six months?

Speaker C: Everything imaginable. So humble software engineers, uh, for uh, front end, back end, full stack AI applied AI engineers, uh, uh, engineering leaders like uh, vp, Head of engineering, um, and also a lot on the forward deployed engineer side which are software engineers that get to spend a ton of time with customers and really be product oriented building out capabilities and new features for customers.

Speaker A: Nice. I don't know if we asked this in the beginning. Is uh, is your all's culture, uh, are you um, in person? Office remote hybrid. What's the uh, set?

Speaker C: We are five days a week in person at the office in San Francisco.

Speaker A: Nice. What's one thing about Servos culture that would surprise uh, a candidate considering an opportunity with you guys?

Speaker C: I think people are surprised when they, when they come on site with us just how fun and nice the team is because it's, it's a pretty intense group of people and uh, pretty elite from, from if you just like looked at everyone's backgrounds, um, you know it's, it's a pretty incredible group. But then when you come on site, um, I think the experience everyone has like wow, this is like a really nice kind group of people that I want to spend time with and, and that is something we want to keep up for sure.

Speaker A: Nice. It relates back to the, the hiring philosophy question. So exactly what's a, ah, lesson from your, your previous roles that's stuck with you when, when building Serval?

Speaker C: I have two so uh, My previous startup, um, I optimized for so many things except for product market fit. And so I felt like the reason why I wasn't seeing crazy traction was like, oh, maybe I'm like, you know, I'm not uh, good enough at leadership or enterprise sales or like I just found all these things to optimize around the edges and it just fundamentally wasn't product market fit. And being at burkata and seeing that you can get so many things wrong if people just love what you're selling them and still be very, very successful, I think that's the biggest lesson is, is what product market fit feels like and knowing that when you have it so, so little else has to go right, uh, and everything will still work out.

Speaker A: Well said. Uh, what was your dream job as a kid?

Speaker C: I wanted to be a wildlife zoologist. Uh, and I actually went to college to study animals. And then I realized I think I more wanted to be like a TV wildlife zoologist, uh, like a celebrity wildlife zoologist. I didn't actually want to go and study some random population of animals somewhere. I wanted to have a TV show,

Speaker A: first time answer on the show. Nice. What's uh, uh, an activity or hobby that you lean into when you're trying to decompress?

Speaker C: Um, it used to be a lot of rock climbing that's been come really tough with uh, a two year old because she uh, doesn't quite fit in the harness yet. I think she needs like 3 more pounds on her and then we can get her in the harness. Um, so I mean skiing when the season, uh, when it's ski season and then I, I've, I've gotten really into scuba diving as well. Um, I wish I had more like everyday hobbies that were really easy. Uh, but I don't. So it's basically like I don't, I don't decompress until it's uh, you know, end of the year or something.

Speaker A: Nice.

Speaker D: Sorry to interrupt. But you can, so you can climb with, with a child?

Speaker C: I. Yeah.

Speaker D: I'm gonna have to see some photos of this. Yeah.

Speaker A: That's wild.

Speaker C: They have full body harnesses, uh, so you just like hoist them up and, and you know like you can, you can pull them and then you can uh, there's all these like, you know, straps on the harnesses. So then you can just like strap her to you.

Speaker D: Very uh, cool.

Speaker A: I thought I was being adventurous putting her in my daughter in a backpack. You know, you're, you're climbing, climbing mountains with her. That's awesome. Uh, what was your first car?

Speaker C: 19, uh, 99 Jeep Wrangler.

Speaker A: Uh. Oh, nice.

Speaker C: I didn't surf, but it would have been a great surf car because you can also.

Speaker D: That would be a great surf car. It would look very cool pulling after

Speaker C: the break in that in South Carolina. It was like ugly, like beige colored, but I thought it was the coolest car. And uh, yeah.

Speaker A: Nice. Folly Beach. What was your beach in, uh, Kila

Speaker C: was my beach in Charleston. Yeah.

Speaker A: Yeah. Oh, nice. Uh, favorite pizza topping?

Speaker B: Sausage.

Speaker C: Uh, nice.

Speaker A: And then, uh, yeah. Charity or corporate philanthropy that's near and dear to you?

Speaker C: Yeah. I think in general, like wildlife conservation, that this is my, like, uh, this is the, the career I didn't go into. So if you go to our office, uh, I've got all this like wildlife photography around the office. It's very much like safari themed and so, um, that's been an area that's always important to me and something I'd love to get back to one day.

Speaker A: Nice. That's a wrap, guys. Jake, Pat, thank you both for joining us on the, uh, on the episode. Great conversation. Excited to keep tracking Serval and uh, the disruption that you're doing in the space. So thanks for joining us on the podcast.

Speaker B: Awesome.

Speaker C: Thank you so much.

Speaker D: Yeah, thanks for having us, Sam.