Name: Finance Forward 2025 - From Data to Decisions: AI Insights at Scale
Uploaded: 2026-01-22T04:06:03.581Z
Duration: 50 min 8 s
Description: Finance Forward 2025 - From Data to Decisions: AI Insights at Scale

Transcript for "Finance Forward 2025 - From Data to Decisions: AI Insights at Scale": We were saying that we've got kind of an unfavorable time slot, like the post lunch solidly mid afternoon energy low. So we're going to try to keep it interesting, And we appreciate all of your engagement and participation beforehand. So yeah, my name's Libby. I lead finance at Clarity. Clarity's been around for seven years now. That whole time, really focused on improving the work of accounting and finance teams through automating document processing. And then within just the last year, we've got a new product that we launched that helps leaders understand how work is getting done in their organizations and really how to think about the AI journey that they should embark on and where to start, where to find savings, where to make improvements. So with that, I'm joined on stage by three wonderful panelists who are at very different places in their companies and careers. I will let them introduce themselves and share a bit about what they do, where their companies are, and then we'll launch into discussion. Good. I guess I can start. I'm Joel. I run the finance organization at a company called Decagon. We are two years old, and we make AI agents for customer support. So if any of you have ever chatted with, you know, Rippling, Notion, Figma, even on chat, email, or in some cases, voice, you're actually chatting with one of our AI agents in a lot of cases. So, very excited to be here to chat about our AI journey. Awesome. Good to see everyone. I'm Michael Gordon. I'm happily unemployed right now. But, just stepped down from almost ten years, working at a company called MongoDB, where I was CFO and COO, and helped grow the company from 40,000,000 in revenue to 2,000,000,000 in revenue. Took the company public in 2017 and have advised private companies. I'm on the public company board of a company called UiPath that I joined prior to the IPO, where I'm the audit committee chair and helped take that company public. And so excited to be here and compare notes with everyone. Great. Hey, everybody. Scott Buxton. I am the CFO of Supabase. We are a series e company that's about five years old. We are a back end database mainly for SMBs and startups. We like to say that 65% of y c graduating companies build on Supabase. We're trying to nip at the heels of MongoDB Atlas. Got a a ways to go there. We're just south of a 100,000,000 in ARR. Prior to there, I was at Datadog for five years both pre and post IPO. Awesome. So to kick things off, I think we're all eager to deploy AI successfully in our organizations and to see improvements as a result of that. One of the things that we've already heard mentioned today is the importance of the underlying data in your organization really as a prerequisite for having any luck with the AI deployment. Could each of you talk a bit about maybe the state of data in your organization today and how you're thinking about focus, improvement, kind of what it means to you for where you're trying to go with AI? Sure. I mean, guess I can Yeah. We're the youngest ones, so it might be the shortest answer. You don't have all that inherited baggage. Yeah. Exactly. So Decathlon, we're two years old. At this time last year, I think we were probably maybe 10 or 12 people. And today, we're right around 200. So there's been a lot of growth and maturation that comes along with that kind of growth and journey. Now, of course, as you can imagine, us on the finance team, it's me and one other person. So the data infrastructure is very new and nascent. Right? And so really, obviously, job number one, it's like, be the data warehouse. You need to get data in there. You need to get it structured in the right way. Same thing from getting the right tools and systems and financial stack in place. Right? And that for us is really the prerequisite that we need to have and where we've been focusing a lot of time in my first six months is get those basic foundations there. So if don't have that, you can't skip to the shiny fancy tool and AI agents doing all sorts of wonderful things if you know, they they don't have the tool to interact with or you don't have the right data in the right place and the right shape and form, for the rest of the company to or for the rest of your workflows to work to work off of. You know, that's rapidly changing and obviously a key focus area for us as we kinda continue to grow. And it all starts for us with go to market and kind of work backwards from there in terms of prioritization on what matters at this kind of stage of a company's journey. And maybe, Joel, a follow-up question. Are you driving a lot of that from your seat? Are you engaging with peers across the aisle in IT? Do you guys have some kind of center of excellence that's helping you with that decision making and that improvement? Yeah. I mean, a lot of this is partnership. So we didn't have an IT person until a couple, months ago. And so for us, it's really across both biz ops and finance. It's kind of a core area of partnerships. And, you know, as we start building out our data team, you know, we're all kind of working working together. And ideally, a story I've seen happen sometimes when a company is growing really fast is this actually happens in silos. Right? It happens in each different functional areas. Those teams get built up at different rates of speed. And now if you're growing fast enough, sometimes every function is growing fast enough that you can do it in a more coordinated way, which is what we're trying to do. But again, it's, yeah, it's doing the hard work of building the relationships with your peers and, making sure that you're trying to make some coordinated decisions where you can. But we're too small to have, like, a center of excellence. If that is you know, I think Ryan's here, who's our controller. I think he's probably the center of excellence for a lot of this, you know, stuff and work. So, you know, when you're as small as we are, I find how it how it works. Yeah. One thing that certainly matches my experience. One thing I would maybe add to that is I just think about it is especially in the early stages when a company is growing rapidly, there usually isn't a deeply established data culture, data environment. Often what I've seen is different departmental functions have a desire to go solve whatever their need is. I'm in sales or sales ops. I need to solve a thing, so I'm going go try and build something. And I'm sure this won't resonate with anyone out there. But like it's possible that team has maybe more budget dollars. And so it's easier for them to go hire a person or find a way to go get something done. And one of the things that I found over time is the best way to tackle this even before getting to AI, but just sort of setting up that infrastructure, is having a team that is at the core, what we used to call centralized but embedded team, right? And so it's not the best, people do it all sorts of different ways, but what I've found is it's not the best practice to have an analyst in each of the teams. You have the CMO goes and hires an analyst and everything else because what winds up happening there is all of this infrastructure that has to happen. And it's incredibly complicated and cumbersome to explain the evolutions in that infrastructure to all a widely distributed or dispersed, not even team, but like set of individuals who are responsible for things, as opposed to having a centralized team, but where they are deeply fluent in their end users business, right, and their end users needs. And that's at least what I've seen, how you kind of get the best of both worlds, right? I'm context aware, I'm not just sitting, answering, responding to tickets, And I understand the business. When someone's asking this question, I actually know what the reason behind the reason they're asking the question is. But I also am intimately fluent with all the data and all the infrastructure in order to be able to get them the best possible result. And I and I found, you know, at a prior company, was weird thing about the first, like, data people to bring in. Right? I know, like, Brad earlier was talking about the second hour was a data person. Right? And it's like, oh, did you just hire someone that does BI or not? And for us, it was always a search for these kind of unicorn type people who are both willing to build data pipelines, understand a little bit of data engineering, but are also able to do some of the analyst work Sure. As well. And it really takes at a startup, it's the unfortunate reality. You need to find those people in a lot of cases. Or if you can find those people, it can pay tremendous dividends down the line. It might not be your gold standard tables, but they'll at least push you along in a well structured way. Mhmm. Yeah. We saw that start at the series c where, to Michael's point, there wasn't a dedicated team that was cross functional. And so we called them the growth engineering team. And, just think of them as billing engineering. And where they really started with trying to wrangle the data and get our arms around everything was with the billing platform. And so we implemented Orb. We're a usage based billing company and so with the subscription component. And so we didn't really have a lot of insight on our customers usage. And so that was the first thing that that they did. And this was about a a year ago. A year ago we were seriously. And then the next thing they did was a BI dashboard. And so we had found that with a 110,000 paying customers, you really get to the limits of Excel really quickly because that doesn't even include your free users. And so we picked Hex for that. But I think, again, to Joel and Michael's point, having the dedicated team that is technical, that's going to work with finance, going to work with everybody else in the organization has been a game changer. It's really hard if you don't have that. Have you seen the perspective or focus on data shift at all in this age of AI within your companies or Michael on the boards or anything like that? Yeah, would say in general, yes. And I think it's probably two trajectories. I think within a company and its arc, as you scale, what does data mean? How do you use it? What does it look like? Evolves with the company. And certainly, you move that out of the vacuum of an individual company, but then think about, all right, what's happening in the broader market context? Certainly, today, people are expecting or fantasizing about much more than they can get from their data. It depends on sort of what you have. And I think Joel's point at the beginning of there isn't very much. Was a different time. But like I was remembering when I first joined MongoDB, there were literally zero people in FP and A, even at $40,000,000 in revenue. I was far along in the interviewing process. Had signed an NDA and I got to sit down with someone who was closest to the numbers. I was like, show me the board book and let me see all this. And we got through the first set of questions and I was like, all right, so what do the unit economics look like? And they're like, well, we don't really have those. I was like, okay, we can create them. Like, what do the cohort curves look like? He's like, we don't have those. So I was like, all right, so what do we have? Right? So it's a little bit like the cobbler's shoulders didn't have any shoes. You were here building a database. We didn't actually have any of our own data. I think it's one of those things that like you just, you think about building a product, you think about sales and marketing, if you're scaling rapidly, you're just sort of fighting the hand to hand combat of daily life. And you don't always stop to think and sort of put those things in place. And so I think you've got to, if you can invest the time upfront, it will become incredibly valuable to you. And then as you scale and grow, you can use that for more thoughtful and intelligent decision making, right? I used to talk about internally with my teams this idea of once we have the data, the burden is on us to have an insight, but it's not good enough back to like what types of analytics talent do you want. It's not good enough to have the insight. You need to be able to translate that into action, right, and operationalize it and then have it affect what you're actually doing in the business. And then you can continue to do that. And then as you evolve into a public company, you're starting to think now about, okay, what are the metrics? Not only do we want to run the business on internally, but how do we want to explain our story publicly? How do we want to be evaluated? What metrics, the context of the IPO, what metrics do we want to share only one time and make it really clear to investors that we're only sharing this once versus what are the metrics that we're going to share quarterly or annually? And then you're going to be measured and having a little bit of foresight to say, how might this metric change over time? And will it always move in a favorable direction? Or do I need to maybe not try and use that because it's favorable now, but it's going to change because certainly investors don't like the goalposts. And then lastly, yes, certainly with AI, once you've got your house in order, there's definitely an expectation that you can use the data more efficiently. And to some of the conversations that were on the earlier panels, leverage the data, automations, AI to have less headcount to be able to grow more efficiently and all those kinds of things and make better decisions for sure. And one thing I found very tactically that I've done, when you're really, really at the early stage and you're like, there's no data. I don't know anything. There's no metrics. To your point, no cohort curves, no unit economics, no operating model. It's really just you start. And for me, it's usually like a meeting. It's like, okay. Let's just figure out how many customers do we have. And then the next one's like, okay. How much pipeline are we creating? How is that converting? And every week, you just kind of keep adding the next metric, and you keep going further and further and further and further down. This is a way of operationalizing things, which is, I know, a very far step away removed from AI. But then in the AI world, right, like for us, there's kind of like the classical world, which is like, you need all the highly structured data to be able to do your machine learning on or predictive analytics or whatever it might be. But there's also the kind of super messy LLM world, right, which is much more of the take unstructured information and actually be able to still derive insights out of that. And that's somewhere where we're still able to do. I wouldn't say so much more like financial analytics, but still a lot of getting understanding, whether it's in contracts or other areas of our business, much more quickly without having to go through the rigor mortal of getting data and structuring it really well. I mean, maybe shifting gears a little bit. Both of you have alluded to it. But where do you start? And how do you ground your team from a metrics OKR perspective? Are there a couple tactical areas of focus that you guys are creating goals around and saying, you know what? We're going to start here this quarter, this year. What's your perspective there? Yeah, I could start there. What we decided to do is back to having so many customers at small ACVs is we wanted to use AI to measure our largest 500 customers. And so, you know, to Michael's point, how can you do that efficiently and effectively so that you can just gain insights very quickly on those 500 customers? And so that's where our growth engineering team built this really product or platform that you can just go and ping it anything on these 500 customers. What's usage as of today? Are they on a committed plan? What are their billing terms? You know, when's their renewal date? And so that was where we really started is just at the highest level of who's paying us the most to try to use AI to measure them. And can anyone use that internally? Or is that sort of permission aware? Yes. Anybody can use it. And I think for us, when you're operating from a blank slate, it really comes for, you know, what is the key goal for the business? Right? For us, it's driving new, you know, new business ARR. It's like, okay. Well, we need to understand that. Right? Where do you go then? It's like, okay. We need to go to Salesforce, make sure it's structured, make sure we have sales stages, sales process. We know what the gates are. We know how things are moving. We're starting to drive some degree of rigor in that, as little as there can be. And you just kind of keep peeling the onion as the business keeps evolving. And you kind of end up just going further and further and further down until eventually, you you actually have a full picture of unit economics. Right? And it's just kind of like left foot, right foot, you know, continuing to peel that, you know, every day, every, you know, every week kind of kind of thing. But it's a it's a slog. And then you change definitions. And it's like, all my old data doesn't work anymore kind of thing. It just keeps growing, evolving, and changing. Yeah. I think the only thing I would add is starting at the beginning and with OKRs is objectives and key results, right? And so you need to understand what are you trying to measure. And I think to Joel's point, being really crisp on that is incredibly important because you can come up with numbers for like almost anything and you can plot it on a graph or see is it trending up or down or whatever. But like they're not all equally important. I've had a number of friends over the years who've worked in or for private equity firms and they talk about their monthly dashboard reviews of 180 KPIs. And I was like, I think you're missing the fact that the K is for key there. They don't all equally matter, And like many people, I was sort of shaped by my very first business experience. And coming out of college, I worked at Procter and Gamble in Cincinnati doing brand management. And P and G is many things. One of the things it is that it excels at structured linear thinking, right? And so there was this very rigorous framework for the objective, what are we trying to achieve? The goal, how are we going to tell if we're achieving the objective? The strategy, how are we going to get to the objective? And the metric, which is how are we gonna measure for achieving the strategy. And then what happened is the strategy for the company then cascaded down and became the objective for the business unit. And it's all worked away all the way down cascading and interleaving all the way down to like the individual brand. And so if you don't know what you're trying to achieve, it's really, you can waste a lot of time on data and metrics and everything else. And then I think at least for us within the finance organization as well, like I find early on, there's usually a series of just more project oriented things that you need to get done. It's not so much like a metric. One, it needs to be something that's also within your team's control, like the point around the cascade. So for us, it was really we need different systems in place. We need to migrate, whether it's on the ARAP side of things. For us, in q two, picking a new general ledger was a key thing for us. That was going be the foundation for everything in the future as well. We're obviously very public about picking Campfire, which we're very excited about. I was going to ask. Yeah. Yeah. Put them through seven hours of demos, I think, is what it what it what it was, which thank you everyone for doing that for us as well. But there's always a mix of, like, MBOs. And then, you know, from my seat, always at a at a start up, you know, cash is really important and being able to monitor cash and make sure you have at least some degree of control over your cost structure is important. So I always try to have a little bit of focus there from an objective perspective as well. Yeah. Say the other thing that's maybe onto your broader question on measurements and OKRs is we went through a number of iterations in the context of MongoDB around OKRs. And we can set aside the whole like, are they good or they're not good, whatever, but like we certainly use them somewhat regularly. And in the early years, it was very helpful and I think disproportionately important for people to see their work and how they were doing it, how it tied back to the OKRs, right? What is that, of the five things we're trying to accomplish in the finance and operations org, how do I know that the thing that I'm working on is tying that? Like, do I know I'm doing important work? Right? And so what we found over time, though, was as the organization got bigger and more complicated, we probably spent a year or two where we sort of have like umbrella headings or kind of tried to find a way to nest all these individual objectives or smaller team level objectives under like some unifying theme. And they kind of got watered down and weren't as helpful or didn't provide as much clarity. And we got to a point where I both got feedback and sort of, you know, I think correctly read the room in that people were mature enough to say, not everything's equally important. We're now a team in the hundreds, right? We were going to say, these are the most important things that we have to get done. And yes, of course, a bunch of other stuff has to get done as well. But let's provide that clarity and visibility to everyone so that if someone is blocked, if the we use the term DRI, Directly Responsible Individual. If the DRI is blocked on that, or it's asking for your help on something, you know it's one of the most important things. And so you're actually going to swarm and help to go accomplish that. Obviously, it puts a lot of pressure on you as a leader or as a finance leadership team or as an executive to be really clear about what the things are, to hopefully pick correctly and then articulate it correctly. But as we got bigger and bigger, we found that actually was more helpful and obviously required some level of organizational maturity for me as an IC or a manager to look up and say, wait a second, I don't see the thing I'm doing that's super important to me. It's not reflected up there. But the answer is that's Okay. We're trying to do a lot of things, but not everything's equally important. And then how would you, within your finance team itself, like would you how would you keep people kind of motivated or try to drive that connection? Yeah. We basically, we cascaded it down. And so my all of my different directs would have their own things. And then we would find ways to reinforce if there's something that maybe was important but not important enough to be in the top ones, where we would set up the use of your time is incredibly valuable. So to your point about setting a project in a meeting, if the answer is, oh, Michael's going have a meeting on this topic, even though it's not one of the top things, clearly what I'm doing is There are other ways to signal or to convey the value to people or to get them engaged and motivated. And obviously, of that takes your own team levitating and elevating themselves. That's how we do it. I guess, are there specific outcomes that any of you are within your orgs or the orgs you're close to, Michael, really trying to drive with AI in the near term? Call it an OKR, call it a metric, but a thing, an outcome that they're watching for, that they're hoping to see or already seeing some change in as a result of where we are right now? So I would say yes. I'll go to mini soapbox diversion just for fun. So AI means a lot of things to a lot of people. And it's obviously the thing that most people want to talk about all the time. I would just distinguish between what I think of as sort of general AI, not like actually in a literal general AI, but sort of like AI legacy AI, I would have called, as opposed to generative AI. And the reason why that distinction is important is many of you run teams or functions that are responsible for an incredibly high degree of precision and clarity. And you care about the controls. And you care about the accuracy of things. And I think we have to step back and understand the probabilistic nature of generative AI versus the deterministic nature of many of the processes that you all will run. And so when we talk about using AI, I think it just gets conflated into this giant buzzword soup. And I think it's really important to be deliberate about what do you mean and let's make sure that it's applicable to the things that you're doing. And so I don't think there's enough precision around that right now. And so yes, we have, even before ChatGPT was launched and everything else, we're using machine learning and data science models to do retention prediction and evaluating cohorts and all those kinds of things, product stickiness, the benefit multi attribution touch models around customer acquisition, or what's the benefit of getting a customer to use a third product, or all those kinds of things, I think, is super helpful. That's different than generative AI. Generative AI, I would say, at MongoDB, we probably saw more success not in finance, but in other teams, including other teams that happened to report up to me. So with my COO hat, one of the teams that reported up to me was our customer engineering organization. So that was all of our technically oriented employees who are customer facing. So I think presales, professional services, and post sale technical support. So we got huge gains from generative AI with our post self technical support team. Because here you've got a mission critical product, right, database. You've got to follow the sun model twenty fourseven. You're doing handoffs from the West Coast to Australia. If you're the technical support engineer and you're picking up the case and you don't have all the context, you have to read through all the tickets and every the history. Whereas Jenny, I could summarize it for you, give it a sentiment rating, and you can be much more effective and much more on top of it. And so we think about customer service, customer satisfaction, of meantime to resolution, all those other things dramatically help. But less so, at least the teams that I'm familiar with, with using generative AI in finance. And I would say for us at our stage of business, there's a lot of different conflating factors that we have. And so we actually, ironically, being like an AI native, we sell agents. We don't have any specific OKRs that are just tied to an adoption of a tool or use AI for this specific thing to move conversion rates or cycle times or whatever it might be. For us, you know, there's still a lot of we've to get basic process in place. Right? Like, we just need the right people in a lot of right seats to start doing the basic job itself to start. Because whether you have agents in places or other quite frankly, for agents in a lot of cases, they still need to be managed. They are a living, breathing thing in a lot of cases. And so whether the underlying LLM models change, whether your own internal processes change, whether your product shifts, whatever it might be, they continue to need to be worked with. And so for us, it's still much more just trying to drive the metric and then leaving it a little bit more to various teams to still figure out what is the right tactic or mechanism. Is that a tool to put in place? Is it a process to put in place to make sure that we're driving towards those outcomes? And so in some areas, again, to your case, a lot of it's more outside of finance. We have a lot on the engineering side of things as we think about how we build our product. Because obviously, we're an agentic solution. A ton on the go to market side of things and a ton on the post sale side of things as well for us, whether that's creating content, whether that's how we think about using various other software solutions within our different processes. Yeah. I think for finance teams, that's where it it is really a challenge because you wanna you wanna use AI to be more efficient, but it feels like the tools that are available aren't quite maybe as popular or prominent yet as some of, you know, the other departments tools, like a, you know, a code generation, you know, cursor or whatever. And so what we've done is we've and the finance team specifically is just said like, well, why don't you make a goal to try something with AI, you know, once a month? And then come and and sit down with us and share, you know, what, what you've done. And like, for example, this last board meeting we had last week, we used an AI tool to help generate our slides. And it didn't it didn't go very well, but it was it was an experiment. Right? Slides of the meeting. Did you did you present them? No. No. No. No. No. It was it Exactly. You know, it came up with something interesting of like, we wouldn't have phrased it this way or whatnot. But I think that's that's a challenge that we see at least is that there is all this opportunity, but within finance specifically, it feels like it's a little harder to adopt. So I don't know if you guys are feeling that as well. I feel that way. Yeah. I would say, specifically, generative AI, I agree. The best use cases that I've seen in finance are for sort of prep or research, right? And sometimes we write it like a first draft. I mean, I think in the early days of ChatGPT, we fed it our earnings scripts and like ask it to generate like the next version of the script, where like all of a sudden it like hallucinated and talked about it like a big acquisition which we'd never done a big acquisition. So like I don't even know like where it came from. That wasn't a good idea. Yeah. No. I think it was like x x y y corporate like it's like not even like a thing. It wasn't actually an M and A idea. It was just like a part of our announcement. But I do think when you're, if you're building a business case, like what are the upside scenarios? What are the downside scenarios? Using it to prompt some of those questions to help you think about, oh, like hadn't really thought about that. Maybe there's this other synergy. Or maybe there's this cost benefit. Or what are the angles I haven't explored this topic? Just to prod you out of your regular thinking. I think, again, so often you have too much to do. Your teams are under the gun. You're trying to get things done. But I think on some of the more strategic projects, using it as sort of like a thought partner sounding board, I think, can actually be pretty helpful. And my guess is 85%, 90% of the stuff you will have already thought of. But there'll be a good nugget or two that are incremental. You're like, oh, that's a connecting the dots that I hadn't thought about or a different take on things that maybe expands your thinking a little bit. Yeah. Yeah, we've lived some of what you were talking about, Scott, where it's like Anthropic announces Clog for Excel. You download it, and you think, oh my god, it's going build the model. And then you try for a half a day, and it doesn't work. But you can use we've got a T and E GPT that we created, which saves me a bunch of time, people reaching out with, can I expense this? Or same for business justification documents. It's like, hey, every time someone comes to me with spend, I want it to look this way and have these categories and have an ROI justification. And now we give them a GPT that helps them put that together. So I think there are projects and point solutions where it's useful. But it's hard to find that kind of at scale fit today. Yeah. I had asked the females at finance group on WhatsApp just what do you all use AI for? And I think it it was a lot of these pointed use cases like it helps me draft an email or it helps me, you know, review a contract. One lady said it automated the creation of a 100 invoices that used to be done manually. So think there I think there are these use cases. I mean, I I just feel like that's where the first panel today was really interesting in the sense that it feels like, again, specifically for finance, it's just making you more efficient. Like, you know, it's getting that last maybe 10% there that you mentioned, Michael. Think we're pretty far away from it to say, okay, like it's going to replace five jobs completely and create our model for us. Yeah, maybe the other last best practice that I would add is the answer is not to identify all of your current processes. Not every process should be kept or automated or identified. I think sometimes it's helpful to question the process in the first place. And is this, even before getting to the automation or identification of it, like, is this actually a process that I should have or I should run? And certainly in the context of scaling MongoDB, just even as I think about myself, I had the number of times that we changed, like the threshold that I approve things Yeah. You know, over that time frame, like, you know, had to change radically. And some of that was excellent. Some of that was a little uncomfortable for me. But it's all part of, you know, growing. Yeah. And I think for us as a as a startup and one of the big things we were really excited about Campfire for, right? Again, I was very fired up watching the last demo because I'm like, oh my god. Was actually like, when you get the right tooling in place, there actually are all these use cases that can exist that will hopefully also, as we think about the scaling of our teams, hopefully get sublinear scaling as we go. And so I do think, especially if you're building a company from a clean sheet right now, there are more chances. You have more at bats to do it a little more cleanly out of the gate. And especially when you have the foundational layer, just like the GL, which a lot of things I think are built on top of, hopefully, it gives us a lot of room to run at least on the accounting operations side of things. Yeah. Totally agree. Yeah, I joined about a year ago. And we were on QuickBooks. And it was hugely painful. And all the data was really terrible. And so we made a Campfire buying decision a couple months ago and are just about live. And we really focused there to start because that is the foundation for any of the modeling that we need to do, any of this other kind of data analytics or AI deployment that we would hope to benefit from in the future. Maybe one last rapid fire question, and then we're going to save some time for Q and A from the audience. What are you excited about in this age of AI? And that can be big or small, near or far away, but yeah. I think for me, it's the promise of agents. It's probably because I think about agents a lot, mainly because I also work at an agent company. And yeah. Well, I just think it forces it's at least really forced me to step back and think about different workflows and think about different areas of the business. But also, to your point, do we even need this process that we have? And as we're really early and we're starting to build processes, it actually gives you the moment of pause to be like, oh, could we build this in a slightly different way? Like, could we think about this approach in a very different way? Like, I have my rule book or my run book that would normally be like, Okay, hire this person at this time, put this tool in at this time. Like, you just can kind of go through the full stacks. It's a known surface area of things that need to be done, so to speak. But that can kind of be questioned. So for me, one of my core tenets of every time I thought about a job or an opportunity or like a company to invest in or whatever it might be is follow my curiosity. And so for me, there's a lot of excitement in this kind of moment that you can have those opportunities to think things through in a slightly different way. I think there are a lot of things to be excited about. If I had to pick most, I think I'm going to jump to the unknown unknowns. Because I think there are a bunch of sort of things that you can identify as potentially being interesting. But at core, I'm an optimist. And so I think there's a bunch of stuff that we don't know how positive it's going to be. There will be some negatives for sure. But I'm going to take the upside on that. Yeah. I think the exciting thing is there's just so much available out there in, call it, vendor choices. Choices. Like, we're on QuickBooks still. We're pretty immature for the size company we are. And, you know, a few years ago, you would probably graduate to NetSuite or some other thing, and there there wasn't a campfire that had, you know, the functionality that was gone over earlier today or, you know, some of these FP and A tools. We just implemented ramp. So we're really early on in that journey, whereas, again, in prior roles I had at Datadog, GitHub, and whatnot, you kinda just went with the large dominant player in the space as you scaled, you did what everybody else did. And it's not the case anymore. So I think that's what's exciting. Yeah. Just add to that because I think it's incredibly exciting. Think it's like Libby and I were talking about this a while back. But having sat in the job for many years and approved lots of software, there were a lot of new hot things in the CMO tech stack. There was always some new sales ops or rev ops tool that we justified replacing the thing that we just put in two years ago, and all of these other things. And if I look at the breadth of innovation across kind of the CFO, office of the CFO tech stack, there's an enormous amount happening. And it should be very exciting for all of you as professionals. And the one thing I also try to remind myself too, as I think through it, there's a lot of amount of experimentation. And with that, to me, comes a lot of discomfort. You also, especially, I think, sitting in the office of the CFO, so to speak, be comfortable with a little bit of failure, a little bit of experimentation. And that's like, Okay, that's going to happen. But the most important thing is actually rate of learning. And actually, I find for most people, just trying the damn thing. Right? Like, just try a workflow. Just just do it. Right? Like, don't don't overcomplicate it too much and just, like, take those baby steps. And that that I found for me. Like, I spent a long time being like, my god. Design is perfect workflow. I gotta learn how to code with these, like, AI coaching agents to do all this crazy stuff. And in the end, it was just like, just start with a simple thing. Like, just write just write a damn email. Right? And just and then slowly from there, just like, for me, it just like layered on top of itself. And it's kind of that beginner's mindset that I found has been very, very exciting, but also nerve wracking. Well, and especially with the rate of development today. Because the same thing that we're talking about is a frustration right now. I mean, it may only be two weeks from now that we would be on the same stage talking about this phenomenal thing that some version of AI can do for us. So we've to keep at it. I think we have some microphone runners. I'll give you all a minute to think about whether you've got a question for us or not. Yeah. Yeah. How do you all think about platforms as a solution? I'm curious, especially to Joel, you're an agent world. And then also, Scott, you're still on QuickBooks. Like where I'm wrestling with is a little bit of like you go the full everything that solves everything for everybody but probably solves more than what you need or do you go agent route and try to plug in agents at specific things and let that scale over time. Hope that makes sense. I think so. I can kinda take take a crack at it where it takes me. So I the both frustrating answer is I would say it depends a little bit, right, both on the the tech stack that you have. And to the earlier point, where AI is a continuum. And you see in a lot of places, they're just embedded classical machine learning type applications, which are fantastic. You can get those in your point in a lot of the kind of solutions that exist today. Yeah. I think at this phase of a market cycle, it's pretty hard to get, like, a generalized platform. Or it's pretty hard to pick one that I think will be durable through time. Like, I think there's so much that we're learning. There are so many different approaches. The models are changing so fast that it's hard to be like, okay. Well, this is the one agent framework that is going to exist. Right? Like, for us, we've changed how we build our agents probably three times, and the company is two years old. Right? And so I say that in the from the perspective, like, probably more like point oriented solutions to just, like, solve the problem that you have. Don't worry too much about building the, like, solution that is going to solve all all problems. Right? It's like, as they come up, TM down, are you going create some technical debt later? Probably. But the world is going to look so different that that's, I think, the right bet to take. Yeah. I mean, I think a platform is kind of a CFO's dream. Right? You saw the demos earlier. You had a tax engine, an FP and A software, you know, billing engine. What if there was one company that could do all those? I mean, that's probably what, like, a SAP would say they can do at the highest enterprise level. But I feel like at your earliest stages, you're still just gonna look for the best in class for what you need it for versus vendor consolidation on a platform. Yeah, would just add, I think platforms in the eye of the holder make sure you as the customer actually find the value in it and not some vendor telling you we have a platform, therefore it's better. What do you think the, ideal CFO candidate profile looks like over the course of the next two to three years? If you are on a board of a company going through a CFO search, what is the talent and attributes that you'd be looking for? I have a lot of thoughts on that. Super hard to generalize. I think specific to this topic, I think what you would call in HR circles a high learning agility is very important. I think there may be a whole bunch of specifics about the company in terms of like, what's the nature of the business? What's the balance of the executive team? Is it likely to go public? How much storytelling do you need? Capital raising? Mean, they're like a very long long. It's really hard to spec out. But I think to some of the comments from Joel and Scott, broadly, I think it's having a business oriented mindset, ultimately being a partner to the CEO in building the business, thinking about, again, I usually think storytelling is important as a company scales, assuming it needs some level of capital. The CEO may be very good at that, so maybe that isn't important. And I think an operational partner, right? Someone, I once got the advice very early on when I first took the CFO job of the idea is to try and create the news, not just read the news. Now you're not at the word meeting saying, here's where all the metrics were. But how do you know where they are, know where they're going, anticipate them with enough action to shape them so that they're different because of what you did? Those are the things that I kind of care more about. Anyone else? We've got maybe one and a half questions of time. All right. No. I saw one over there. With all the AI tools that are out there that you want your teams to experiment, how are you balancing that with the security and IT concerns as well as the financial burden that all these different tools might have? Yeah. Mean, for us, we put them through a review. And it's the usual balance of risk that you think about, like what data is it touching, where is it getting used. There's a few different inputs for how you can think about risk from that perspective. If it's an internal tool that looks like customer data, yeah, we think about that a little bit differently than if it's something that is just maybe more like the marketing world kind of thing. So you have to kind of take the usual risk based approach for thinking about it. I think it's also depending on what your business does and then also what stage you're at. Think your risk tolerance shifts and changes. Yeah, I would I would add, having fewer bright lines, but having some clear bright lines about sort of the dos and don'ts, right? Like just what on your assessing your risk posture at the outset, the other thing I would say is I would just strongly encourage sandboxes for early experimentation. But just trying to think of what are the critical things of how could this go wrong? Just making sure you're protected in those most extreme cases, but then try to encourage you to iterate and innovate as much as possible within that. Now we'll let you go. Thank you, and thank you to our panelists.