LaunchPod - Dave Ruda === Dave: [00:00:00] For every response we get to these automations we scan all the inbound email . if you and I were emailing in a personal thread, at some point during the conversation you might say like, , let's get together on Thursday and Siri today she'll go, would you like me to schedule that into your calendar? We did the same, we did it for collectors. It's a functionality that reads all the inbound emails once we realize there's categorization, this is a promise to pay, they're logging a dispute, they forgot their invoice, it'll categorize that so the collectors can now focus on the different activities and priority. collectors usually take around eight minutes per email. Just in six weeks we've been able to drop that to two minutes, 30 seconds. Jeff: Welcome to Launch Pod, the show from Log Rocket, where we sit down with top product and digital leaders. Today we're talking with Dave Ruda, VP of Product at Billtrust. In this episode, we'll discuss how Billtrust built their new collections agent, that four x to capacity for thousands of collections teams, their decision framework for identifying where AI can add the most value for their customers and why Dave sees human in the loop as a crucial piece of FinTech AI for both trust and efficiency. So here's our episode with Dave Ruda. Hey Dave. Welcome to Show Man. Thanks for jumping on with us. Dave: Thanks for having me. Jeff: Yeah, dude. [00:01:00] Yeah, it's this'll be a fun one. I'm glad this worked out. You know, background that you're bringing here is pretty cool between what running product right now over Billtrust. But before that, you worked at a company called Take Two that made video games and worked on quite a few ones that people may have heard of. A little indie game called Grand Theft Auto. Dave: Well, you can't gimme all the credit. I mean, I don't, I don't make the Jeff: You didn't make it. You worked on it. Dave: yeah, we, we work on the infrastructure around it. Yeah so I've worked at Take Two Interactive it's the parent company for Rockstar Games and 2K games. So NBA 2K GTA Red Dead Redemption. And my favorite SIV six Civilization. Jeff: You're basically naming a lot of my activities from college right now, Dave: Partially the reason I didn't finished engineering school. We'll just leave it at that. Jeff: Let's jump into the main thing I wanna talk about, I was in New York earlier this week and we had a whole dinner and we had about 30 product executives who were there and we were going through kinda what are the problems you're facing? And one of the major things. That people kept saying like, my board or my CEO or some stakeholder, more senior to the product team was basically saying, , you need an AI [00:02:00] strategy or you need more ai. And it was great 'cause in this case I had the one-liner that was my new favorite thing from you. Which is basically you can either salt Bay, a little AI on it, or you can actually do a real AI strategy. You lived this actual situation not too long ago, your board actually said you guys need an AI strategy but you didn't just Salt Bay it, you actually went back and created, software , that. Drives real outcomes. Now that I've kind of set the context for that, you can tell a story way better, let's dive right in. What's the thing that's going on here? Dave: Let's address, like, why do boards say that? Jeff: There's a real reason like it. It's legit. Dave: every company out there who's looking for software solutions or just to level their game up, they're looking for the angle that's gonna make them differentiated to make their people more powerful. The other analogy I like to use is like Ironman, their employees, like Tony Starks steps into the Ironman suit, becomes a superhero. He has all the great gadgets, to be able to do it. And he could do like 10 x more than what he's normally doing. it's not a weird ask, , it's saying , how can we be more modern? That's really what you're asking for. And then the second part of that directive is there's a kind of a sales angle to it. If you're [00:03:00] gonna go and purchase something, you want confidence, right? You wanna feel good about the purchase that you're making, so why would I get in bed with a company that's not using the latest and greatest? I think that's really the question that they're asking, because there's an aspect of this that is marketing, straight up, Hey, we use ai. Okay, great. Then after you say that sentence, it has to really mean something. And that's where we started. , You might get the directive that's great, but the best way to approach it or the way we try to approach it, which may be the best way to approach it, is where do I find pain and. When we say ai, now everyone knee jerks into GPTs 'cause that's the hot note right now. But the fact is, like AI's been around for a long time, Billtrust particularly, we were doing AI before AI was cool. We have machine learning integrated into many parts of our app. Our cash applications rife with it. We use business rules, which is also ai. And that doesn't make us antiquated. It makes us solid. There's good fidelity and result that you get from these functionalities, and they're tried and true. The real question when , you get the directive nowadays is like, how can you GPT things? So what does GPT do? , It takes unstructured data and it builds structure around it where you might have to write really [00:04:00] complex algorithms or code or Java or whatever the code base is to be able to kind of wrangle that solution. And then create a great outcome. And there's finite walls there, whereas the GPT can kind of think on the fly. Dave: The real question they were asking us is, how can you address some of the pain points in the process where customers see this pain and then be able to address it because the technology now exists Jeff: similarly like. You know what GPT-3 hit in 2023, I wanna say. And that's kinda when the whole world went bonkers over all of this. We've had a team similarly working on AI since 20 17, 20 18. But it was, you know, machine learning back then and we always had the vision of, this is where AI is going , and we're gonna get there. And a lot of that work was. A distilling signal. It was gonna be useful at the moment. And then B, getting things ready and building towards a world where the tech caught up to what we wanted to do with it. This was not like a overnight, suddenly outta nowhere, AI has been around and I thought I was cool that we were able to say 2018. Dave: These are not new concepts. We talked about me in gaming, I worked in hospitality before that, and travel. We were [00:05:00] doing, large scale analytics We're looking for is what's the utility we're trying to drive and then how do we take the right tool off the shelf , to apply it? And I think companies will just create bots. Which is great, but where's the use? And I think the thing that we're seeing is when we think about these solutions, we try to put them into the human in. Yeah, the human in the loop situation. And a lot of companies out there advertising human on the loop. But , human in the loop situation is we're in this, fetal stage of trying to adopt these functions and, lots of jobs or roles within companies, , the question I always ask myself when I'm using a new tool is, do I trust my job with it? Like, would I be okay if I was wrong? Because if I'm wrong, maybe I don't have a job anymore. We work in finance, the FinTech space, so we're moving large sums of money here, and you gotta be able to trust that. You're making the good choices , we don't take that responsibility lightly. So when we try to introduce functionalities, not only do we do great extensive amounts of qa, but does the functionality that we bring forward, does it allow me to trust that with my job? , Am I putting the faith in the machine? So the human in the loop factor is really important right now. And the other thing, you touched on it a little bit, was that, 2020, what is it? 2023 was like the first [00:06:00] time we saw this new model. This is really a good moment to be able to release these kind of capabilities because now you look at MCP servers are available now and you can create good rag models and CAG models and these are new protocols that existed now, did not exist before to create that learning loop in this unstructured way so you can build good coaching models and that's what allows us to drive really great outcomes for our customers and also create cool shit. Jeff: And Maybe let's take 10 seconds just for anyone who doesn't know, , you're at Billtrust. Clearly it's in FinTech, you know, the ten second Dave: Oh yeah, the elevator. Yeah, Jeff: yeah, Dave: Yeah. So we're an AR automation platform. We accomp many companies out there. Jeff: accounts receivable for the non-finance. Dave: right, right. I got, you're gonna have to get me on the TLAs, the three letter acronyms today yeah, , we're in the accounts receivable space. Large companies, B2B companies, they'll have an ERP, which is their books. And the ERPs have base level functionality, nine times outta 10. Things like sending an invoice, manual processes, maybe some light collections software inside of it, potentially a cash application software. , The wheel of the different functionalities and processes [00:07:00] within, inside of a accounts receivable function within the finance department. But the ERPs are often lacking , the scale and size and the workflow that's needed to be able to do this. So , we're both onto that. We added additional value to allow you to scale your business. Without having to scale your staff. And , that's the way we like to say it because we're trying to grow every year and all these companies are growing and they're getting new accounts, they're sending out more invoices, they're trying to get paid on those invoices, and we wanna make that process as seamless as possible, but also give you the best chance to get paid by, by driving the right outcomes, not just for your customers, but also for your employees, so that they can get the money in the bag. More often than not, what we see is that the finance department is usually the last modernization of an organization. It's a necessary function. It's often seen as a cost center, and it's just did you do all the things? Because , if everything gets done, then everyone's happy. If something goes wrong, then everyone's very sad. Right. It's like being a system officer or something like that. Jeff: No one notices you until you mess up. Dave: . Exactly. I think about our positioning as us being a revenue driver. So long as you can drive. The right outcomes and figure out like where the friction points [00:08:00] are. So a good example is like setting credit lines. Sometimes they're set and then they're forgotten. They never revisited. And these are great ways to drive revenue opportunity. If you see repeatable patterns like good payment history good payment on term, not delayed, not delinquent what we often see is, , and this is true for any software, it doesn't matter if you're in FinTech or not, there's usually a decision maker who's going out there to buy the software. Great promise, great ROI Sounds great on paper. And then they hand the project to somebody who's maybe never done a protect project before. Doesn't feel empowered enough to do the change management, the top down ownership or just not enough experience to understand the art of the possible. when we design our solutions, we're trying to create things that. Are attainable for the persona of the end user, so that it's almost like their process folds nicely into our system. Don't get me wrong, we've had some customers who come to us and try to take the crap process, put it into the cloud, and they go, we did it. Mission accomplished. But the reality is like there's a lot of sadness going on under the deck where you're just doing the same process over and over again. In the design process, it's not just like, now that we have these tools the unstructured capabilities, even the machine learning capabilities, [00:09:00] how can you present the facts in such a way that it forces the adoption? then also is like integrated enough into the process that it feels like a natural extension. It goes back to like that Iron Man scenario, right? Like , if you're out there fighting the MUS for matter space, like you don't have time to figure out like the user manual and how to fly the Ironman suit. It's gotta just be so intuitive and natural part of my process that I can, win the battle. Jeff: And so you guys kind of took this approach looked at, okay, what are the things we can solve? And it sounds like the first thing you landed on was. Collections. Every business has a small amount of revenue that they're not collecting on. They have, accounts receivable that just Dave: You say small, Jeff: into, well, it depends, you know? Dave: It's big. It's usually pretty big. There's, yeah, there's a write off usually at the end of the year and it starts with seven digits usually by, regardless of the company. So there's a lot. Yeah, there's a lot of money out there and there's a lot of gold in those hills, so, Jeff: So, okay. You, so you picked accounts receivable only collections, which has a lot of money that people don't collect. , , if you can make that better, that's a gigantic gain for you and your customers. Dave: the whole thing is a scale problem. So imagine you work in the financial organization, AR department, you're a collector. There's only so many tasks you can do in a day. First you have [00:10:00] to triage the account. Understand. How many bills have they paid? What's their history? Are they in good standing? This is a one-off. This is a common occurrence. Like you have to first distill all of that, and then you have to send them the reminder. We call this the Dunning notice. It's like, Hey, you owe me some money. The way they respond usually falls into a couple of categories. Like, oh, I didn't know that I owed you that. I thought I paid it. Can you send me the invoice again? Sometimes it's like, I already paid that. Why are you yelling at me? Or Hey, I'll pay you on Monday. Normal response, right? imagine now you're that collector and you have a book of a thousand different accounts that are overdue. Most companies , you're trying to grow your company every year, so you need to have a certain ratio of collectors to the amount of accounts that you need, manage. And you can think of it like a CRM if you're not constantly contacting the right people at the right time, , and you're doing this through a manual process where like you go into the ERP, you download your aging buckets, they call them. So like how many are 30 days? Supposed to 60 days, supposed to, and they just work what's overdue. But there's a philosophy that we have around early intervention. If you can know ahead of time, and this is actually not related to the feature we just released, but you can [00:11:00] machine learn out each account into where you should focus, we're releasing a feature like this in Q3, so maybe I'll come back on the show. But it can prioritize which accounts. And then the feature you release is really how do you optimize the outreach to make it easier for the collector to response. They can touch more accounts and if you can touch more accounts, then you get better collections. Jeff: that's number one is just who do you even reach out to, it sounds like. And then you have , this agentic process for collections, which is now how do you. Execute on this. 'cause at some level there's, having a human being, the sole thing doing collections , is tough. 'cause humans scale tier wise? So as you grow you need to keep hiring more humans to collect more revenue outstanding. And so , you're always having this kinda like skim on your margin. , And if you don't do that. You start to see more uncollected. How do you look at that problem? What's the kind of tool look like , and why were those the capabilities you chose to go after here? Dave: if you're sending reminders or dunning notices or making phone calls, those can all be scheduled and automated for the most part. Like, hey, today's the day you gotta send the automation, and the computer could just send that out. That's not an issue. You probably get them from your medical bills or you get 'em for, paying Verizon later, something like that. That's a pretty normal thing. But the more [00:12:00] complicated part of the process is what happens next? So how does the customer respond to you? That's where we saw most of the time spent. Besides phone calls, , it's harder to reduce the amount of time just because you have to have a conversation with somebody. But in this autonomous way, how can we make it a lighter lift so that the collector on the email threads can digest this? And also, you gotta imagine the swivel sharing, right? You and I might be talking and then, Steve down the corner might email me and then I have to completely change my frame of reference, reread the entire email thread, , and for most of the people probably listening on the podcast today you're probably gonna think about like your day at work, right? So you get an email from the marketing department. Certain context , then you have a finance email. Completely different context. You have to completely shift your frame of reference. And some of these threads can go really long, right? I'm sure everyone at work has like a 50 thread, TLDR, you've popped that in the chat, JBT, and you're just like, all right, I came in late to the party, just summarize this for me. that's what we wanted to hone in on. Those are the hours that we can get back, right? You don't wanna remove the personalization from it, but you want to understand like, one, what's going on. And then two, if I know what's going on and I have enough information to be able to take an [00:13:00] action, let's take the leap to that, There's interesting and there's actionable and people will pay for actionable. So for every response we get to these automations, the reminders, the dunning notices, we scan all the inbound email correspondence. You know, sometimes, like if you and I were emailing in a personal thread, I'd be like, Hey Jeff, how you doing? And you'd be like, I'm doing great. And then the GPT would only know that Jeff's doing great, but at some point during the conversation you might say like, Hey, let's get together on Thursday and Siri today. If you're using Apple products, she'll go, would you like me to schedule that into your calendar? Right. It's a pretty normal thing. So we did the same, we did it for collectors. So someone like Jeff might say, Hey, I'll pay you on Monday, and we go, that's a promise to pay. Let's log that and then we'll follow up with Jeff on Monday if he doesn't do it. So what our agentic, we call it agentic email. 'Cause everyone's gotta have agent in the title now. But it's a functionality that reads all the inbound emails and then once we realize there's categorization, meaning this is a promise to pay, they're logging a dispute, they forgot their invoice, they want to talk to you about something else, it'll categorize that so the collectors one can now focus on the different activities and priority. So [00:14:00] that's a really good win already. Like, Hey, let's lock out the promise to pay us today. Those are low hanging fruit. I could just get scheduled. The customer's already making me , a bet that they're gonna pay on a future date. For the customers that are missing automated invoices, I can bang those out too really quickly. Disputes, maybe I gotta deal a little bit longer on those, so maybe I'll work those third. But you can see your stack ranking, your priority of the different actions already. And then when you get into the email, we can pull out all the unstructured data. Those are things like, , Jeff said he wants to pay me on Monday and he wants to pay me three grand. So it's gonna say, log a promise to pay. $3,000 knows what day today is and knows what day Monday is contextually, and then it's going to fill in all the blanks. And instead of them manually going through and trying to parse out the whole email and figure out what Jeff's saying, it's gonna take that into a single click action. obviously there's hallucination out there because, , maybe Jeff says things a little bit differently than Dave and we still create that human loop experience where you can trust the verify and you can change all the fields too if you had to. So it's single click with confirmation, and then it logs the action and you move on to the next task. Now, net net we see collectors usually take around eight minutes per email. Just in six weeks we've [00:15:00] been able to drop that to two minutes, 50 seconds, or, sorry, two minutes, 30 seconds. So yeah, big time savings, four x return, four x as many touch points Jeff: Right. Dave: made. Jeff: five and a half minutes doesn't sound like a lot, but when you multiply it over, how many of those emails they're touching and how many actions they're doing, and the fact that you can, make that team far more efficient is if they can do three or four x more actions in a day, that means that that's, , one third or one fourth of people you theoretically need to hire. In that role, which adds up really, really quickly. Dave: What we see is like, people go, oh my God, are they gonna replace me? Like, that's not the answer. The answer is, , we just want you to have more meaningful conversations where they need to happen. Jeff: I have yet to see this thing come to fruition where, people have successfully got rid of entire teams because of ai. I have definitely seen that you can work on more interesting things and maybe it hasten you kind of like looking at in reality what's priority and what's not. But generally I've seen almost every time it ends up being the people don't get replaced. , They get to do something that's less drudgery work Dave: what was your first job outta college? Jeff: I wrote marketing copy for , a publisher that worked in like tech [00:16:00] publishing. I wrote descriptions for white papers , and summaries for white papers to put into emails to try and get people to download those white papers because we were the intermediary between companies like ca. And the end user. So basically we published those and promoted them and promised we'd sell them that we will get you a hundred leads for this white paper. Dave: It's a little repetitive after a while. Right? Jeff: just wrote summaries. It was, mind numbing in, in doing the same thing over and over and over again. Dave: The reason I bring that up, , my first real gig post-college was at an email factory. Like we would create emails and we were managing lists, and I would, look at large Excel sheets and copy column A to B and move the data around and access and so on , and sql it's like mind numbing stuff. After a while you're just kind of doing the same rote memory right? think about that from a collections point of view. You're kind of doing the same thing all the time and the persona of the person is, somebody who is probably an hourly worker in some places or they outsource it, so they're doing this re repetitive motion. I think the retention on those roles, it's painful because they get tired of the role. They walk out the door, the knowledge walks out with them. You gotta retrain. They get burnt out. So, this is a good opportunity to improve your retention, right? You want people to focus on really interesting things [00:17:00] like why do people stay at jobs? They have a really great manager. They're learning something, right? And they're doing things that are meaningful to the mission, right? So this really hits on that last one, like you're really focusing on areas that you know are gonna drive the best outcome. Tools should compliment that. So it's not just about like, yes, the net net of it is like, let's get really great collections going, which will happen, but also you wanna like coach your people into this. A fast follower to this because the technology exists is we're building a learning loop into the system so that we look at the initial generated response. The collector edits because it's editable, right? Like you want it to have your voice and you could do that now. Like I talk differently than you and you want them to feel like they're not talking to a robot. So with MCP servers, you can do that. And now we are building this learning loop in. So all this great training data that we have from the first six weeks of launch that's gonna be coming in this quarter to create what I call like the voice of the collector or the voice of the supplier. And we're building these in a framework way because we know this is not the end. in fact, we already have a plan for Q4. We're gonna be releasing it with voice. So we're gonna be able to pull out all the information from the transcript as you're having a phone call, and allow you to [00:18:00] pull out those tasks too. So like everything you can do in our email solution, you can do on our voice solution as well. Jeff: Not taking them outta the process, but just taking the drudgery out of it similarly, what we're doing here, we started out doing session replay and analytics. But the problem in product is what are you gonna do? Watch 10,000 session replays , even if you somehow get that data magically, you have your customer feedback data in one tool, you maybe have support tickets in another tool. You have like eight tools you have to go across, and each one of them has a piece of the story. How do you get the entire thing altogether I've literally talked to people. Who spent hours doing that and talk about it being three weeks between a support ticket coming in and the product team finding out about it. So , we built a solution that connects all of that, and also the agent that watches the sessions. So that it can basically pull out, you know, you released something yesterday, we saw, here's how it's doing, and here are the issues. Or like, Hey, this ticket came in to to Zendesk or Intercom. But we noticed that only one person complained, but that same problem happened to a thousand other people. You should probably pay attention to this. And how do we service it up? It's not getting rid of PMs, it's allowing them to not spend. Hours and hours of drudgery parsing through all [00:19:00] these tools, but rather, how do you spend more time with customers and how do you spend more time solving the important problems? Dave: , The amount of output now is astronomical. So I think about it like, just creating like one feature, one deck about one feature. To pitch to, a board member, an executive leadership person. I would spend like three weeks on that and I had to manually pull all the data together. And then I probably would miss something because like, realistically, I'm not a data scientist. . I'm a human but now it's like you don't even have to have those skills. You can just. Put the information in it finds like the, I don't enough about data science and the theory to be able to like leverage the GPTs in my daily workflow to be able to find those anomalies. So when you think about the tooling that's out there now for PMs is just incredible. We should be creating better products, , stronger outcomes, better trends, finding insights if you're really leaning into lean and agile, which we try to, and, and I don't mean that like, oh, we're lean and then it's waterfall as hell. We are trying to actually feedback all that in. So like what you just mentioned is exactly the approach that we try to take. Now we have the tooling to be really like supercharge it and make it for real. And that's why things like that [00:20:00] learning loop came about we've even redesigned the UX a little bit before even a couple weeks after because of the great feedback loops that we see in this data and how the machines can pull that out for us. So it's a super exciting time to be a pm I think there's some PMs out there that are like, oh no, the GPTs are coming. am gonna be able to write PRS anymore. Oh no, not that. I'm like whoever liked writing A PRD to begin Jeff: I, I've said this before, but I don't trust anyone who enjoys writing a PRD. It, it was a necessary evil, but there's better ways to do it now. Like you said, I, I think one of the things that really impressed me when we first talked was Not only did you like, it's very easy to go like, okay, we're not going to just sprinkle a chat interface on this thing and move on. But it was a very thoughtful approach to like. , Where do people get value from? What our software does? Where do people run into problems or where is the friction? What are the high impact areas where we can really drive a lot of good here? And what are the tools available to us? Really good at not just like use the tool to use the tool, but like, where are they best deployed, Dave: I love being out on the field with customers more than anything. And if any of our customers are gonna be watching this, I want you to know that you'll see me soon in many of the events that I'll be out there talking about this. [00:21:00] So, and, and or on different phone calls that I, I typically take every week, I start there. We call it day in life, like you sit over the person's shoulder who's doing it, and you try to understand that Then you look at your product telemetry and you try to understand the friction points and like does the qual match the quant? Right? And then I do something that a lot of product leaders don't do. I say, what? I buy this for those that do work at Billtrust and especially those in their sales organization are gonna cringe when I say this is I bought a shit load of software in my career probably more than anybody needs to in their entire lifetime, to date, and I've sat across the pitch desk from like really great salespeople who can articulate it and those that have not. And I try to build solutions and value prop it. like I'm selling to myself because I think I'm a pretty harsh critic. Especially now, timeliness is that every dollar is being scrutinized within all these organizations. So a dollar out is the equal 10, 10 x back. That's just a fact. Jeff: It has become wild now. Like what used to be a hundred people in an organization could approve a purchase. That same purchase is now, there's like three people that can sign off on it and they all have C in the title probably. Dave: Exactly, [00:22:00] so , you're putting bets on like what's going to be a winner and like as a product person, if you don't feel like this, then you're probably not working at the right organization. It's like you're making bets out there you better get a result because those r and d dollars, they're all counted too, and you gotta make sure you're getting 10 x return. It's like gonna Vegas, like I would love to know where the advice are gonna fall or which number's gonna come up on the roulette wheel. And if you have enough points like using the quant and the qual to do that, that's good. But then the last set of testing is your gut check. I'm the harshest critic of my product 'cause I fully own it. You're the mini CEO of your business. , It's like an extension of who you are, what you're bringing to the market. The best capabilities are always born out of people being the harshest critics of themselves and trying to figure out those edges. So I think, I think we're onto something here at Build Trust we're hard on ourselves because our customers expect us to drive the best outcomes, and I constantly try to put myself in their shoes. So I actually think that , the functionality of the AI and, all these bets that we're making. It's gotta follow the same pattern of what creates the best products, and that's listening to your customers, putting yourselves in, in their shoes. So product's not really going [00:23:00] away. It's really like finding empathy with the customer and then applying the right outcome using the right tool Jeff: Done right , the leverage of product , will go up here. Looking at our, recent release one of the big ways we honed it was we went and talked to a ton of customers. About , what are your problems? Where do you spend time? Where do you spend time usefully , and non-use usefully? And , this part about, oh, it takes hours to answer a question, I have to go through six tools and it's all different data all over the place. You suddenly start to go, hold on. You aggregate data, you kind of like pars. It is all basically text or easily customizable. It's just a matter of, of collecting it and understanding it in, context of each other that is something , that not only can , we build a tool for , and save you hours of time in a day, but we also have this central source of what we also can distill and have the agent watch sessions for you and tell you what they experienced there and make the data you'll get from the other time even better. But we can also remove that time. So now summarize that other stuff for you. Put it in context of this summarized session experience data. And now you really have a great insight. But the last part that comes from it is getting people to use it. The thing I loved about how you [00:24:00] guys looked at it was put it in the flow, right? They still have to approve it. They still have to check it, so they feel a part of it. The change management there is , very natural. But you have to deliver these things in workflow, right? Dave: that's the mentality right now. I don't want you to think I'm saying that fully autonomous will not come. It will, I don't think the technology's a hundred percent there yet like, going back to my original thing, like, would you trust your job with this? Like, we're not there yet. we are looking at things like going fully autonomous with what I would call like the low hanging fruit. For forgotten invoice, do you need or a missing invoice? Do you need to have a human do that push button? Probably not, right? There's, that's low risk, high reward, more time back. as we look into 26, a lot of what we're gonna be focusing on is , how can we create AI governance so that corporations still feel in control that the right tasks are being handled with the right utility. Right now everything is human in the loop. We wanna be able to create a new zero, right? Like the new base. And that would be everything below this line should be handled by a robot. Everything above this line needs to be handled by a human, and I think that's where we're gonna evolve and eventually the line will keep going up. I think that's how, as like the GPTs evolve, that the technology becomes [00:25:00] more durable, flexible has better learning loops scalable, then we'll be able to do that , at a greater pace. Now you mentioned like only two years ago we hit GPT-3, right. And now we're at like, I think 5.0 came out today, I think that we're gonna see , a big change over the next six months, probably in the next year. We think about Moore's Law and the processors and how the gpt are being adopted. It's just a really fun time to be in product. Jeff: you know, linear, , the dev, ticket tool. Everyone knows linear, , it's hugely popular. They published recently their agent interaction guidelines. One was basically an agent doesn't actually own any action like you as a human, you may delegate to the agent, but you still own the outcome. How do you think of that in kinda relation to some of the more fully automated areas you're kinda looking at down the line? Dave: In our space, there's a lot of new entrants they go fully agentic, fully automated right off the bat because it's, they're like, let's be sizzle. Ah, so what we're seeing is a lot of the entrance into the space are building for the fully autonomous, and it's funny 'cause they're getting wins. Like we come up against them in deals sometimes and like they, they'll get a win. Okay, cool. [00:26:00] But, and then people go, oh my gosh, we have a new competitor. Go fully Auto-Tech. And I'm like, whoa, whoa, pump the brakes. If you notice, when you look at what they're building, they're actually building backwards. So they started with that. Now they're building all the foundation to build the human visibility to on the loop and the visibility that's needed to be able to feel like the company's still in control It's like self-driving cars. I guarantee you, like people are gonna be weirded out when they get into a self-driving car from the beginning because you're sitting there in the backseat and you're like, okay, well this is how I, this is how I die today. This is how it's gonna happen. , It's the same thing, like, I don't think emotionally we're ready for that. Especially when you're saying things like, would I trust my job with this? It will happen over time. Just like, when the iPad came out, everyone was like, what's this thing? . Who's gonna buy this? And now, you know, touchscreens and tablets are everywhere. It's just like part of life. Jeff: On the marketing side, I, you know, 'cause I, I sit and market, I work with our product team all the time. And I'm involved with product, but I, I run our marketing team and I run the SDR team here, and there has been this huge move similarly to like fully automated ai, SDR, prospecting, all sorts of stuff like that. And I've seen a ton of companies that went to use it and then a ton of companies that moved away from it. Because what they found is [00:27:00] without the kind of . Step that you've described of human more in the loop and being able to kinda control the actual bit more. You saw maybe some early results, but it ended up doing some things that they were really unhappy with or the results just weren't there. Or usually Dave: Oh yeah, like Delta and the airline. Like you could, you could say, gimme a free ticket, and then all of a sudden everyone had free tickets on Delta Jeff: Oh, I didn't hear about that. No, no. But you know, this is, I think just more like subtle abuse of just over emailing customers and things like that and that kind of thing. But same thing like all these startups came out and got tons of traction right away because , it sounds sexy, And then you actually use the thing and it's not, and that's where like, you know, you guys came at it from a what is the real problem? Let's solve it. We have, where do we have deep expertise to, to really have the data to solve these problems in a really trustable way. It's finance, like you said, you don't wanna mess it up Dave: I appreciate that and I appreciate you having me on today. We talk a lot about like how we're improving ar it's nice to talk about the product perspective a little bit I think what we're doing , is special. We're trying to create tools for people , that are doing something the same every day and we wanted to get them to be excited about coming into work. And we also want them to drive real outcomes and that's really what matters at the end of the day. Back to my same thing like do you have a manager you really [00:28:00] love? Do you feel like you're connected to the mission? That's something that's often forgotten in the finance department. Like, what am I doing that's gonna impact the total organization and how can I show that broadly to the rest of my organization? , And then obviously we want people to come in every day and feel energized about what they're doing, they're learning. So I think we hit on two of the three reasons why people stay. And then also if you're a C-level person watching yeah, we do collect a lot more cash with our Jeff: Yeah, if you make the, team happier and help them make more money, that's holy, holy grail of combo right there. appreciate you coming on, man. This was great, Dave. Thank you for joining us. We'll have to have you on in a couple quarters and see how the rest of this is going because this is cool stuff you guys are doing. . If people have questions or wanna reach out LinkedIn the best place or is there a better place to reach out to you? Dave: LinkedIn's great. Take it from there. Jeff: That's where we met. I think that generally works. Speaking of getting value outta things, if you got value outta this and you like the podcast, you like what you're hearing. If you're on YouTube subscribe to the channel. Give us a, like subscribe write a review please. That is how we get the word out on this thing. Doing that will help us do this more and, if you can tell one person that you liked this that is the number one thing you can do. But yeah, thanks everyone listening. Dave it's a pleasure man. Look forward to Dave: Thanks. Jeff: to have you on again soon. Dave: Thanks [00:29:00] Jeff.