Starting Lean, Thinking Big: Treasury Lessons from Trupanion’s Samma Hollier

Speaker A: Foreign.

Speaker B: Okay. Welcome to Fintech Corner. We're here at AFP, uh, 2025. I'm joined by Sama Olier, uh, senior treasury manager, recently promoted. Um, and I'm so excited about our conversation. And I had to get it out of the way right at the beginning. Um, my wife and I were a dog family. And I'm so curious at a company that is about kind of like the health and well being of pets. Do you guys allow pets in the office?

Speaker A: Oh, absolutely. Um, and there are no rules around it. We have some of the most, like always dogs barking and like, one will bark and then like the next five will bark. Um, and then every Wednesday we have Pup cup day. So you can hear. We have an ice cream truck that comes around and all of the dogs will start freaking out. And so we have like a. Which one's the best one with the pup cups. Uh, we are super, super dog friendly. There are some cats as well. We've got a couple hairless cats that will be pushed around in strollers around the office. Uh, but you definitely, you can't not love pets if you're going to be coming to my company.

Speaker B: Was that just like, uh, when you walked into your interview for the first time, was it just like, yes, I have to be here?

Speaker A: Uh, I don't think my interview. So I, I interviewed during. We were still recovering from COVID so it was all on Zoom. But the first thing, every single question is always, what pets do you have?

Speaker B: Yeah.

Speaker A: And are you allowed to work at Trupanion if you don't have a pet?

Speaker B: Right.

Speaker A: And if you do, uh, the amount of people that have joined and one year later have a pet is exponential.

Speaker B: Yeah.

Speaker A: Yeah.

Speaker B: That is so funny. Well, in. Amidst the chaos of all of the dogs barking and cats walking around, what has kind of bringing to life a, uh, treasury operation, uh, that is more sophisticated because of technology looked like for you at Trupanion.

Speaker A: So I think the big thing for us is that. And it's hard to benchmark as well, because most insurance companies, you're looking at the state farms of the world, the fidelities, they're very large and we're very small and we're very niche. Right. So we are technically property insurance. Um, and so being pet insurance, we're like, we're kind of more health, but we're not health. Um, so I think the hard part of that is we're small. Right. So I, when I, when I onboarded Truvada, I was a team of one. And so A lot of mine is how do I scale and how do I forecast, how do I make sure that we have a simplified bank structure.

Speaker B: Yeah.

Speaker A: While doing it by myself. Right. And able to do it by myself and not creating a mess and not having those gaps and having those controls in place, that has been the biggest challenge of it.

Speaker B: Uh, you know, when, when we designed Travada, it's been at this point seven years since our first kind of like true design session. But from the very beginning, we always imagined what would it be like to enable one person to do the job of many. And that was kind of the foundation of the design and where we put a lot of our effort. And hopefully it's something that you've seen in the design is the simplicity of labeling things, whether it's, uh, like transactions or finding things. Like in analytics places where typically you would have had to jump through a few hoops to get to information. It's intended to feel like Google. And I've been curious to kind of pick your brain on this because, you know, being a lean operation, what does that do for you? And also, as you kind of think of expansion, um, does that change kind of the mindset of onboarding new team members and what they will do even in their job?

Speaker A: Oh, absolutely. I, um, think my main thing is when I first came on, we didn't even have Treasury. I was the first ever Treasury.

Speaker B: Yeah.

Speaker A: Uh, but I was logging into all these different bank portals. Right. And then in order to do anything, I was exporting it into Excel.

Speaker B: Yeah.

Speaker A: So I didn't even have transaction level detail. Right. I would have to click into this portal and if I wanted transaction level, I would have to export.

Speaker B: Yeah.

Speaker A: And as a insurance company, we are paying claims. Right. Claims on the daily.

Speaker B: Right.

Speaker A: We use ACH and we use check. And if you look at a bank and you see paid checks, they are each. It just says check individually, a check. So if I'm exporting, exporting, I'm exporting like a 10,000 line Excel sheet for my checks. And how is that helping me? It's not. Right. Um, and then Truvada gave me this opportunity that I could just be like, oh, what is that total number? Right. Yeah. What does that mean for my cash forecasting? How much quicker can I do it?

Speaker B: Yeah.

Speaker A: I don't remember what your question was.

Speaker B: That was kind of along the lines. I mean, the big thing that I'm always curious to hear about is does that A, save you time? Which it kind of sounds like it does.

Speaker A: Oh, absolutely.

Speaker B: But B, does it change your behavior in any way? This is something that we, we kind of really strive for is like, if your original kind of cognitive load was dealing with automation problems and that is removed, what does it do? Right. Does that put you in a different posture? Um, does it make you more analytical? Does it make you more strategic? Does it make you more creative? Right.

Speaker A: I think my main thing was how do I make this quicker every single time? Like, uh, onboarding. Travada first was like my, my out, like cash positioning took me an hour every morning. And now I literally log in and go to the cash position page and I can just see all my balances right there. And then I think now it's, how do I make that even quicker now? Right. Like, what's the next thing? What reports can I create? What analysis can I do? Yeah, um, all the time I'm thinking about has our claims. Um, we just announced that we're going to bring our Canadian claims in house and I need to create a forecast. And I'm like, how quickly are we building claims when we need to fund our claims accounts? How quickly has that grown? I just go into the analysis tab and I can see, oh, it was 100k last week, but now it's going to be 120. And so we're getting that 20% growth before I would have to export again. And I'm exporting every single line of that check. Right, right. Um, so everything is allowed to be quicker.

Speaker B: It almost seems like it's not just the fact that you have to export, but you also had to kind of build the infrastructure in Excel to do the thing, uh, whatever you might have wanted to do. So in that example, let's say you're doing an analysis of a specific type of check, and you want to find those specific checks. You have to pivot that. Then you want to put it into a chart. Well, let me figure out how to do that. Put it in another tab. And it seems like that process. It's almost as if, at least from other customers that I've heard this from, just the thought of it is enough to dissuade you from even doing it. Because it's cognitive load that's just like, oh, uh, man, I have other things that I need to do. I'll get to this later. Or I'll have like, you know, maybe I'll make an estimate instead of getting the exact result.

Speaker A: Well, and then what are you going to your boss with?

Speaker B: Right, right.

Speaker A: You're like, I spent eight hours today making an Excel formula.

Speaker B: Yeah.

Speaker A: You know, and, uh, versus like now, it's like my boss could come to me, ask me a question, and with Truvada, I'm able to have that result within a half an hour. Right. And that's in a presentable way that I can literally just like, if he wants to see a graphic, I have that. If he wants to see a table, I have that. You know, and so that right there, that is also. I'm a, I'm a manager. I'm not supposed to be building Excel sheets anymore. Right. That's, uh, that's what I haven't. Now I have an analyst. I finally got one. Uh, but that's what she should be doing. Right. And so Truvada takes a lot of that out, you know.

Speaker B: So in the world where you're growing your team and you're kind of imagining the roles that the different members of your team might play, how does a piece of software like Travada, but also data infrastructure that's in place that makes this possible, play a role in deciding what will each of these team members do? Because if automation is just gone as a task, what becomes a priority?

Speaker A: Well, I think, I think my main thing, the main thing that I want from her is thinking, what is the next thing? Right. And how do we improve, improve these things? And what are we, like, how do we actually understand those transactions? Right. Because now we're not spending the time being like, oh, look at this transaction went out. Oh, it's payroll. Because Travat is doing that for us.

Speaker B: Right.

Speaker A: What does that mean? And where are we seeing, how is that changing over time? What different cash flows are coming in? What can be more predictable? Yeah, um, that. And then also we already have all this data. How can we use that data for other teams? Right, right. How can we incorporate this? Like, I haven't done it yet, but the GL functionality. Right. How can accounting be using this? How could tax be using this? What are these other teams that could be really utilizing or even FP and a, um. FPA always looks at their forecasts from that high level. Right, right. They get the budgets and they, they say that that's going to come in. How do we, we have this level of detail that we can see every bank transaction. How do we tie that to what they have? And so that's what I really want is that the data is already there for you now present it now use it to improve our business and take our business into the next thing.

Speaker B: You know what's interesting about that, that thought leadership of Saying, okay, we have solved the automation, we've solved the uh, analysis problems that we had in the past. You're not done. Let me keep taking it further and think about how we can integrate business systems together and make them more productive, make them more insightful. It's so interesting that I've heard that thread a few times now, just even in, you know, the past day at this conference talking to customers about how they have seen their role evolve from one of uh, the just doing the job of a typical treasurer to now almost playing a, almost a data architect role for their finance organization. Because you are sitting on the ultimate data, if you will, uh, from a settlement perspective. And I'm curious, what does that look like for you now that you know you have a platform that's organizing it and it has APIs and it can connect to other systems, what are those conversations look like?

Speaker A: And you know, I'm working on those. I think my main thing is I need to get these other teams into Tribada.

Speaker B: Yeah.

Speaker A: Because before it was just me. Right. And how do you sell this platform that only you're using? And so how do you sell the benefit of it to these other teams? Yeah. Uh, and the more that I'm seeing it, the more I'm realizing I will ask my accountant, be like, hey, what do we think for cash flow? What do we think is going to be available versus restricted? And I could see her log into the bank portal.

Speaker B: Yeah.

Speaker A: And she, she's doing that old fashioned thing. And so how do I teach them that there's this resource that they can be using?

Speaker B: Yeah.

Speaker A: And I haven't gotten there. I'll, I'll be honest, but I want to.

Speaker B: Yeah. I feel like, especially for those tasks that are, let's be honest, boring, you know, they take time talking about, I love exporting. Yeah. If they could just take advantage of something that gives them the exact answer in seconds. I feel like it opens their eyes and it also begs the question long term, do you see the integration of all of these systems as part of the domain of your role, like managing kind of like how data flows to the erp, for example, how data flows to any closing system that you might have, um, how you provide information to the C suite, et cetera?

Speaker A: Oh, absolutely. I think that's one of the biggest changes that I have seen since treasury didn't exist. Right.

Speaker B: Right.

Speaker A: When I joined the team and I've now been there three years. Um, there's always this learning curve where CFOs learn that treasury forecasts are Very, very different.

Speaker B: Oh, cash is important.

Speaker A: Right. And that looking at cash from a Treasury's perspective is very different than looking at FP and a forecast and your global forecast. Right. Because real cash out the door isn't the same as accrued cash. Is it the same as restricted cash? Uh, and understanding that and the hyper focus on. We need a forecast. Right. We need a forecast that is our actual cash and can we make payroll and do we need to take on debt. Right. And are we being strategic with our cash, having a resource to do that and getting it out of Excel?

Speaker B: Yeah.

Speaker A: Right. Because you uh, look at even like fpa, FPA is still using Excel and I think in most companies FPA is using Excel. And think about it, every single thing is a, is a formula.

Speaker B: Exactly.

Speaker A: You delete one formula, your whole thing is off. Right?

Speaker B: Yeah. And a lot of those documents are in the cloud. If a sync error happens, corrupts the document, everybody that's been depending on it,

Speaker A: someone didn't open it as read only. And so someone's going, the new analyst is coming in, you know, whatever happened with um, AWS the other day, you know.

Speaker B: Yeah. No, I mean these things happen and if you depend on one file like that without any support, I mean it's, it's super risky. And I feel like I have said this for years at this point, but that single point of failure could be the difference between you making payroll or not.

Speaker A: Exactly.

Speaker B: And um, I know it's kind of exaggerate. It's, it's an exaggeration, but it kind of isn't. Especially if you're depending on the day to day cash flow forecasting and well,

Speaker A: and payroll is usually one of your largest forecasts.

Speaker B: Right.

Speaker A: So you get to end of month and you forecasted badly and you don't have the cash to make payroll.

Speaker B: Right.

Speaker A: That is literally if you ask like any treasury person, they're going to say that's their. What keeps them up at night is not being able to pay payroll. Right.

Speaker B: So let me ask you this. Where do you want to take your program from here? We've talked a bunch about different items like you know, getting into a more analytical mindset, more creative mindset for or for your analyst team. Um, how you want to connect with other folks in the organization. I wanted to pick your brain. What do you think about the future of forecasting? AI topics that really bring another level of automation that is even further than what we're talking about.

Speaker A: I mean I want it so bad and I think that um, AI is the Future. And I think that there is going to be a way.

Speaker B: Yeah.

Speaker A: To integrate it into forecasting. And there's like right now, I don't think we're quite there yet. Yeah, I think, uh, where I really want to see the AI. I've always mentioned this with Travato, whatever, whatever I'm talking to uh, David, or

Speaker B: hit me with it, I love to

Speaker A: hear, but I think that there's a lot of opportunity in tagging. Yeah, the tagging. Let's be honest, that's everyone's least favorite part. You got to do it up front. Which with Truvada it's the only work you have to do.

Speaker B: Yes.

Speaker A: The only work you have to do is these tags. But I think that there's a better way of doing it. Right. I think that because it's so much data, the human brain struggles to be like this makes sense. This is a trend. Like you don't see those trends as well versus with the machine learning. I feel like it really can be like, oh, this, this kind of ties to this and it makes sense that we could do, do this. Which then once you've got that AI creating your tax for you, then that flows into everything. Right. Then you can be creating your forecast off it, you know?

Speaker B: Want to hear something crazy on that story? So we ended uh, up building a labeling system, an automatic labeling system, four years ago.

Speaker A: And that's what I want.

Speaker B: We just had it in the pipeline and we were testing it. We were testing it and we, we were putting a lot of effort into trying to get to like the 99 percentile of accuracy where we could have a recommender that would open up all of your transactions and basically identify ones that have been untagged and then recommend what should they potentially be in terms of tags. Then ChatGPT came out and we found that we effectively threw away the old system because it was so much better. And what we've noticed other customers do and we're keeping an eye on it, is they're just asking inside of Truvada AI, what should my tags be for the last 30 days?

Speaker A: How are you doing that? I forgot that you had that AI functionality.

Speaker B: Uh, I forget all the time that you can do this because it's so open ended Travada AI, you can literally ask it anything. So I kind of run out of creative ideas when I'm faced with it. And this is something that we've been talking about from a design perspective. Um, because it's so flexible that it's almost too flexible. It's it has almost like a creative barrier right at the front of like, oh, it can do anything. Well, what do I even want it to do? And then you just get stuck and

Speaker A: it's like, uh, well you also get worried because like it is so customizable. Once you get so customizable, how do I pass it along to my analyst? Right, right. You're going to be able to understand what I've done. Yeah. If I've customized to much, this was one of my biggest fears. Studying it myself, I'm like, this is the way I want it done. But doesn't mean that's the way you want it done.

Speaker B: Right. It could be completely different. So we've been messing around with the idea of could the large language model suggest categories for transactions. So give it a hard category that is more typical across every customer and start there so that there's a little bit of consistency across customers. Because one of the things about tags that's really interesting is that you can make it whatever you want it to be in the most absolute fine grained detail from the most high level to the most, you know, low level concepts.

Speaker A: You can make it as the level of we paid Microsoft.

Speaker B: Right.

Speaker A: You could make it the level of AP or you could just be outflow, you know.

Speaker B: Right. So that what we were experiencing was spamming basically. There were so many tags suggested that it made it not useful. And we wanted to figure out a way could we create a starting point that gives you some leverage right out of the gate, you know, without having to sift through 100, uh, 200, 300 tags that it came up with. So we're still working on it. It's actually an interesting problem because AI has made it a lot easier. Um, and it can find so many details that if you just sat down and put it into Excel you could come up with a lot of these ideas yourself. But because it's automatic, it just gives you so much leverage. So rest assured, we're working on that for sure.

Speaker A: I know more AI. I'm always very pro AI. Um, anything that you could incorporate. I've been going to some of the AI things here and it is very cool. I mean it kind of, I feel security also though. Right? Yeah. And I love that you have the product within the system. But, but other things that we've seen, I've seen like helps you write policies that you can upload your policy into it and then it makes recommendations based off that of what your position should be. Right. Should you uh, repay debt, should you uh, Invest in money market funds. Should you go longer term? It's very cool. It's definitely up and coming. I don't think anything's coming from my job yet, though.

Speaker B: No, I don't think so either. Because at the end of the day, the treasury role is a risk role, and risk cannot be calculated uniformly across every single organization. Every single organization has a different appetite for different reasons. And I think that's why, at the end of the day, the role of the treasurer is to basically mitigate and manage that risk, you know?

Speaker A: Absolutely. And there is a lot of risk.

Speaker B: Right.

Speaker A: And how are you thinking about all of it?

Speaker B: You know, I wonder, just thinking about AI from a different context, do you feel like there's ever going to be a time, you know, if you fast forward five years from now, ten years from now, and we really start to take the handcuffs off of the systems as they stand today? Because even in Truvada, as you mentioned, security is our highest priority. And data, uh, uh, you know, governance is also our highest priority, which means that we put significant handcuffs on the AI systems. It cannot ever leave your instance. It can't do certain things, uh, intentionally. Um, we don't let it forecast, uh, uh, you know, uh, beyond a certain point. We don't let it give you advice on treasury policy. We don't let it, you know, make payments on your behalf, uh, initiate FX trades, etc. We don't let it do a lot of things. But in a world where we inch closer and closer to removing those handcuffs that are intentionally made, do you ever see a possibility of, you know, in addition to having an analyst or two on your team, having a team of agents on your team?

Speaker A: Oh, man, oh man, that scares me. I mean, I feel like, yes, right. There are certain things that eventually I could see. I could see it from the simplicity of like, um, if you have a centralized, like, like the internal banks. Right, right. Like you keep all the cash in one and you figure out like the working capital of your international entities and kind of running cash and at the end of the day. But also there might not be an end of the day anymore with right time payments. Right. So I'm like, maybe not, but I do think that I could see it. Maybe they are a whole team of agents. I hope not. I hope not.

Speaker B: Um, it's so interesting. Uh, I. So yesterday I had a conversation that blew my mind about how this could potentially play out where I didn't even think of this. Where you could imagine one agent, for example, is focused on fraud and all that they do all day, they. They're not real. It's like, it's. All it does all day is literally look through every single transaction one by one and assess is this fraud or not? And if it is, flag it and then give you a notification if it is, and then if it is, do something about it. And I was like, huh?

Speaker A: I mean, honestly, that's. That seems like a good thing too, right? But then are the robots going to take over, you know, and then are they actually creating the fraud?

Speaker B: That's always the fear, right? And I think that there's got to be a line at some point. But, you know, whenever I start to listen to these use cases, like another one is payments, right? The job of, okay, we have these position targets at the end of the day, by 3 o', clock, the wires have to go out.

Speaker A: Yep.

Speaker B: Who's gonna do it?

Speaker A: You know, honestly, and it would be nice, but like, I also, like. And we. But you worry about that, right? Because there's so much fraud around that and can AI catch it?

Speaker B: Right.

Speaker A: You know, um, and. But like, is it more, Is it better than human?

Speaker B: I don't know, but I, I start to think about it just as another computer program, you know, like, it's so easy to get lost in the, uh. Well, it's Terminator and it's, you know, we're all gonna, you know, experience Skynet, but at the end of the day, it's just code that's running and executing. And while it's a little less predictable than normal code, it still is just code that's executing on an automation that you set up. So I think about that all the time and I'm like, strong with that, like, if it saves me time, you

Speaker A: know, I mean, uh, I was just talking to one of my friends who works in IT though, and he's like, the amount of times that I've had to like, go up to an analyst and they'll be like, well, my Excel is broken. Like, something is obviously wrong with the Excel. And he'll look at it and be like, did you use ChatGPT to do your formula? Because it's dividing by zero and that's why, why it's not working. Um, and so like, also there's that reliance on chat, GPT on AI that we then now have taken out this human error, right? Of like, we. We don't know how to diagnose what went wrong because we're relying on this AI.

Speaker B: And if you start to daisy chain a lot of that where there's one layer of ChatGPT that was put into that Excel formula that nobody checked that might be working now, but not. Might not work two days from now. And then somebody builds on top of it.

Speaker A: Oh, yeah.

Speaker B: And somebody builds on top of that. And then you have to triage the whole chain and try to figure it out. That's nasty.

Speaker A: Yeah. Oh, uh, yeah. But I do think that something's coming. I don't know. I. I haven't heard of anyone replacing treasury folks with AI yet.

Speaker B: I don't think it'll ever.

Speaker A: But I also think that we're asking for, as treasury folks, I think we're asking for more in the AI space than we're getting yet. Right.

Speaker B: We're.

Speaker A: We're always kind of in finance, the last team that these things come to, so.

Speaker B: And you mentioned starting lean, you know, building. Building from being the first. Y. Um, leverage can only come from so many places. Right? Y And if there's no budget for other hires, you know, where does it come from? So it's been a fascinating topic. It's one that I feel like we're not even close to being done kind of sorting through as an industry. But it's interesting to hear everybody's perspective on it. The last thing I'm curious to hear your perspective on is having gone from, you know, being the first implementing systems, now having a team, getting promoted, going through those stages. What advice would you give to our listeners? Kind of going through a similar experience, maybe having to evaluate software here, you know, on the showroom floor, trying to figure out, out of all these options, everything that's going on, how should they even think about the problem?

Speaker A: So my main advice to anyone trying to grow themselves in treasury is networking, um, whether it is with treasury stations, with service providers, with bankers, because especially as a team of one, you're not able to do it by yourself. You're not able to get that knowledge, even if you can watch as many videos as you want. Yeah, you could ask ChatGPT. But when you're really going to get. The real advice is when you grow your network and you ask other people what they've done.

Speaker B: Right.

Speaker A: The whole reason I came to Truvada was I got it recommended from one of my peers and I asked around and, uh, I have so many peers that use all these different treasury management systems. And this is the one that made sense to me because I used my knowledge and then I asked around in my network and I think that's my biggest advantage advice is that even with the AI and all this technology coming up.

Speaker B: Right.

Speaker A: Ultimately, I think you're going to be most successful in treasury if you build that network.

Speaker B: That's amazing. Well, Tama, thank you so much for your time. This has been awesome. And that'll do it for us here on Fintech Corner. Till next time, Sa.