Name: Copy Of Axos Summit Recording FINAL
Uploaded: 2026-03-04T14:09:17.434Z
Duration: 1 h 27 min 46 s
Description: Copy Of Axos Summit Recording FINAL

Transcript for "Copy Of Axos Summit Recording FINAL": Good afternoon, everyone, and welcome to today's virtual summit. My name is Erin Chesterton, and I'm on the marketing team here at OutSystems. I'll be serving as one of your moderators for today. As we give folks time to settle in, please take a moment to introduce yourselves in the chat. Let us know where you're joining from, the organization you represent, and what specifically brought you to the summit today. We'll get things kicked off here officially in just a minute. Hey, George from Snowy NH. I'm in Snowy Maine, so not too far from you. Thanks for joining. Alrighty, looks like we've got a good group of folks in here. With that, we will get things kicked off. As I mentioned, my name is Erin Chesterton. I'll be helping moderate the session today. So please don't be shy. Let's engage in the chat. Keep questions coming as we go. It's great to see you guys joining from all over and definitely encourage you guys to engage with your peers in the chat, not only within the content of the session, but you know what's working, best practices, tips, tricks, whatever you guys would like to discuss as we go. And we'll officially kick things off. It's truly an honor to bring together so many leaders across the banking and financial services space to adopt to discuss a topic that's on everyone's mind, agentic AI adoption, and most importantly, how to scale from pilots to production. We're excited to spend the next ninety minutes with you all in this peer led discussion, where we'll be moving from agentic AI theory to practical application, diving into four real world use cases that Axos Bank has successfully tackled. Our goal is simple, for you to walk away with a clear blueprint to help bridge the gap from AI aspiration to execution. Now that we're all acquainted, we'll cover off on some quick housekeeping items before introducing our speakers and diving into today's agenda. First, today's session will be recorded. We'll provide a link to the recording and the slides within one week following the live event, so please stay tuned for that in your inbox. Second, we will be launching a series of several live polls throughout the session. Please participate in these to help us gauge where you are in your AgenTic journey and your goals for the future so we can better tailor today's discussion. Finally, we want this to be a true peer discussion. I'll be managing the chat as your live moderator, actively elevating your questions to our speakers throughout the session in real time. Here's how to best participate. We encourage you to keep your cameras on for the session to keep things interactive. However, please adjust to your own comfort level. Second, we ask that you please stay muted until your question is directly addressed by myself or one of our speakers for the open discussion. When I or a speaker calls on you, please come off mute to engage in a direct dialogue. As mentioned, feel free to drop your questions in the chat at any time or simply use it as a space to connect with peers to share best practices or lessons learned. And we will reserve any of the unaddressed questions for the open Q and A portion towards the end of today's summit. With housekeeping now complete, I'm delighted to introduce today's speakers and our agenda. First, have Luis Giraldo, our VP of Partner Alliances at OUTsystems, and he's joined today by Kevin Hearne, Senior VP and Head of Consumer Bank Development at Axos Bank. In a moment, Kevin and Luis will introduce themselves and share a bit more about their backgrounds. But first, I'll quickly cover off on the agenda. We'll set the stage by covering the state of AgenTik AI in banking and financial services. Here, we'll discuss the current landscape, including top drivers pushing financial institutions to adopt this transfer transformative technology and, crucially, the blockers and governance challenges that many of you face when trying to push from pilot into production. Next, we'll move into the core of the summit, which is Axos AgenTik AI Evolution. Here, Kevin will dive deep into their practical, real world journey without systems. He'll highlight the four use cases they successfully moved from pilot to production, giving you a tangible blueprint for success. This is where we shift from the theory to the tangible results. Then we'll move into the open q and a discussion. This is your chance to directly engage with both Kevin and Luis and your peers. So, please continue to submit your questions throughout and we'll be compiling them and fielding them directly to the speakers within this portion to ensure you walk away with the answers that you need. Finally, we'll close the summit with a quick wrap up and ways that you guys can stay connected and in touch with our speakers and your peers who joined today. Now with that, I will hand it over to Kevin to share a little bit about himself and his role at Axos Bank. Kevin, the floor is yours. Great. Thanks, Aaron. So a little bit about me. I joined Axos just over five years ago. I came from a background completely outside of the financial industry, heavily focused in supply chain logistics, Six Sigma Black Belt, SAP solution architect made the transition to Axis largely because they wanted, I guess a disruptor you could say, right? Like they wanted somebody to come in and just kind of question the processes as well as look at the technology stacks that we had. And I'll get a little bit more into that in a little bit. Axos itself was founded twenty five years ago on the July 4. And that's a significant date intentionally because as you're aware, you know, markets are closed, banks are closed. And as the first internet only bank, it was essentially setting the precedence of being able to bank even when banks aren't open. Have several lines of business, as you can see here in the slide, legal entities, if you will. We have conventional consumer banking products, checking, savings, mortgage, lending products, things of that. We've got direct to consumer invest products, self directed trading, managed portfolios. We also have a securities side which constitutes our clearing house, broker dealer services, everything in that space. We've got commercial bank, marine finance, equipment finance. So very, very diversified portfolio. And then this year our C suite was honored to be able to ring the closing bell at the NYSE. So pretty significant achievement for the bank hitting its twenty fifth anniversary. Great. Thanks, Kevin. Always a pleasure to conduct these sessions with you. So Luis Giraldo from OutSystems. I run our partner alliance business across The Americas, which is really US, Canada and Latin America. OutSystems has been equally around for around twenty four plus years, focused primarily on helping drive innovation for organizations around low code application development. Now, not all low code is created equal, so we'll talk a little bit about where fits relative not to just low code, but harnessing and building out the future with AgenTeq. And so if you think about AltSystems, Kevin and I were actually kind of discussing this earlier is I like to say it's the best kept secret, but we've been around a while and actually are supporting pretty large scale enterprise mission critical applications across many institutions, across banking and financial services. And one thing is for me to say, another thing is looking at sort of third party analysts to support that claim. So you can look here across the top of the slide. From a user experience, G2, which is well known as far as app dev platforms, mobile development, and overall low code. Industry leader in Forrester Wave as well as in Gartner Magic Quadrant now. I think we just encroached, I guess, plus years of being in the top right quadrant of the Gartner analysis. And as part of that, it's a global community, right, that supports application development for many organizations. So 500 plus partners and industry experts that support that application development using OutSystems to drive that innovation for many customers, and a rich, robust community of developers across the globe. That Kevin has definitely tapped into to support the innovation happening at Axos. And really, think as we talk about where AltSystems moves into the future, AltSystems is leveraging its roots of AI powered low code application development to support enterprise modernization efforts and now providing organizations like Axos to harness that speed of AI and control of low code as a unified platform to support the innovation, but also keeping the proliferation of sprawl at bay. So I think as we think about this, it's not just speed, but it's future proofing and manageability as you're starting to bring agents into your business and starting to drive more of the agentic future. So I'd summarize in a couple of points. AltSystems generates full scale applications across and helps automate the SDLC process, which Kevin will touch upon, helping also create and orchestrate AI agents in a governed way, in a secure and governed way production ready, which is a big theme here from going from pilot to production. And then harmonizing AI generated code with security scalability in that performance and ensuring visibility and observability across the way. So really think about the unified platform approach as the foundation to harness the power of generative AI, which is the topic at hand. So before we dive into the meat, I think, you know, we always like to get a pulse from the audience on some key topics. And so I think many folks have started their journey or have about to start and are trying to figure out where to begin. Or other folks have already commenced that journey for a year or plus or more and are hitting roadblocks. So would love to get feedback from this crowd on the first poll for today, on where are you in your GenTick AI journey? So we'll let those roll in. So thank you for responding. All right, approaching 50%. Alright. We'll give it about ten more seconds or at least when the when the numbers start to plateau. Alright. So let's take a look here. What do we have? Still evaluating at 40%. Ready to pilot. Looking for the right use case to begin. Okay. Interesting. So majority of the respondents are in those first two, which is I think it's a great sort of segue into the session here today with Kevin. Having him walk through the journey at Axos and how he's started the foundational work and now he's really looking at going from pilot to production, leveraging the platform and harnessing generative AI. Great, thank you for the responses and we'll jump right in. So, the state of, AgenTeq AI in banking and financial services. I by no means am an expert in banking and financial services. Kevin can layer in here, but I do work a lot with many of our strategic clients globally, and particularly in The Americas, that fit within the banking and financial services. And so some of the things that I think we wanted to touch upon to set the stage, and I'll love to get Kevin's perspective on this, which is there's no question, right? The AI race is on and we're seeing a lot of evaluation of different technologies out there to bring forward generative AI capabilities and overall just bringing agents into the enterprise responsibly. And so we see that of the banking leaders, 70% feel that it will be game changing. And we'll talk a bit about some of the areas that we're witnessing and where Kevin is looking to leverage that within Axos. And now 51% are already piloting, right? But that being said, I think as it relates to the responses in the first poll, there's still a number of roadblocks and constraints that sit within every enterprise and no different within the banking and financial sector. So legacy systems is always a core component of the strategy. Where do you start? What do you leverage? What do you bring in? What do you have versus build versus buy? I think that's the question. And Kevin can touch upon some of those themes. The unclear use cases in ROI, I think the second point here was what's the right use case? Where to start? And I think what Kevin will bring is a bit of of how do you get started? Show some proof of value. Use that as a segue to determine sort of what can be production ready and ultimately what's going to provide impact to the business as a whole. And then the scaling beyond pilots. I think there's a lot of experimentation that we're seeing. I myself witness every day with our partners and our end clients around, know, starting in internal areas, let's say around knowledge management as an example, even to customer deflection for call centers. So there's a whole host of use cases that we'll share with you. But Kevin can give a perspective within specifically that access where they're focusing on. So to set the stage, this is kind of what we're seeing as some relevant themes and hopefully that are pertinent to the audience here. Kevin, would you like to weigh in on just kind of what you're seeing within the banking and financial sector? Yeah, and I can tell you. So about a year ago when I was looking at just industries and AI adoption, the financial services industry in general only had about a 12% adoption rate of AI and that included like pilot, whatever. And it was one of the lowest, right? Of all industries, it's one of the, it's been one of the slowest to adopt. So to see just in that one year timeframe, this statistic of 51% is obviously encouraging. It means that it's people are realizing that it's not an option. Right? Like, you're gonna get left behind if you don't. So obviously, like you talked on, I'll I'll speak more about how we're using, you know, specifically agents inside of the AgenTek Workbench and OutSystems to not only integrate with legacy systems to improve, you know, roles and processes within our SDLC, but how we're also migrating and upgrading, if you will, in whether that be end of life upgrades or actual full lift and shift. So we'll get into more of that. And like you said, use case definition. And I saw a couple of really good questions. I think Kevin Lang was one of them and George had another one. I'll probably touch on both of those throughout. So I think we'll get some of those answered just inherently within the presentation. Thanks, Kevin. And I think as we get into kind of further expanding on some of the roadblocks, not to stay too long on top level themes, I think it's important, and you and I have discussed this, which is there's always a tendency to kind of lean or over rotate on the shiny bright object. And there's still some core fundamentals from a foundational standpoint, whether it be the platform or the people side, or the governance aspect of this. And then naturally the infrastructure to support that and the data and where it resides and how to leverage that accurately and effectively. So you're not, you're not kind of compounding the whole construct of hallucinations, for example, or other areas that are going to mislead, you know, the perspective. But love to get your quick take on these three elements because I think these are core to any transformation that's happening and particularly now with the disruption of generative AI coming into the fold. Yeah. So on this first point of talent and speed, I'll actually share a little bit more about specifically the Axos journey without systems. But essentially the theme here is that we're finding I don't need to go to market and find highly skilled or experienced developers to get into this. The use cases that we'll go into developed have by very junior people. One has only been with us, well, I guess three years, but fresh out of college before that and completely homegrown on the OutSystems platform. And the second has only been with us a few months and just started learning the platform. I'll speak more about how we adopt and how we get that training done and ramp people up. But this is a common thing I hear from either potential or existing OutSystems customers is how do you find the talent within the platform? And I'm not going and looking for the talent per se. We obviously have vendors that we've partnered with from a staff augmentation perspective where we told them, you know, I told them twelve, eighteen months ago, hey, it's in your best interest to go start building out COEs and out system. The ones who listen obviously have a nice pipeline with us. But outside of that, and by the way, they follow my framework on that ramp up. We basically just gave them the playbook. Now with FTEs, we just do it in house. I've proven that internally my FTEs that we got right out of college are outperforming anyone that we've kind of reskilled from like, say an Angular perspective. Right? So I think it's not as daunting as people may think when it comes to talent. It's not like I always equate it to some of the larger platforms that were out there in my past, where it was like, and I'll cite some like, we were a MuleSoft shop in a former life. And every time I needed a MuleSoft developer, was I'm flying somebody in at an insane rate. Right. That's not the case here. I don't have to deal with that. So that's positive. Yeah. Well, Kevin, is it just interject there briefly. I know before we jump in, you'll touch on this. I mean, I think what I've always appreciated your approach. I think it's a balance of diversifying sort of maintaining some of the internal IP right the team members. And there's going to be a mixed bag of more seasoned traditional high co developers with an entrance of newer, maybe next gen, I would say for lack of better terms, developers that are coming out fresh, maybe more digital native as a perspective, but then also tapping into the partner ecosystem. So I think I'm seeing that as a calibration across all of those in order to effectively support a lot of the transformation and innovation that you're driving forward at Axos. So quick take there. I know we want to get into the meat, but I think it's important to touch on the people side. Yeah, for sure. And I'll speak a little bit more about it, but there's definitely a blend of all of the above. And not all of the high code developers make the cut when you talk about reskilling. Right? But it's not for lack of being challenged. It's just breaking that preconceived notion or the constructs they have and their inherent desire or need to get under the hood, right? For lack of a better term. And it's like, just stop. Like, you don't need to do that. And for those that have the aptitude for wanting to solve more complex problems, those persist. Even in a low code platform like OutSystems, you still have a need for highly specialized JS or CSS developers or backend is always going to be there, right? Highly complex API integrations, things of that nature. And then you get into some of the new features in ODC where now you're getting, you know, how do I leverage Data Fabric layer more effectively? How do I put a caching strategy in place? And so there's always going to be opportunity for those developers that have the aptitude for more than just, you UI work or whatever. So. Yeah, perfect. Thank you. Real quick though, the trust and governance too, I'll get more into that. Obviously there's different ways to approach it. We've had some learnings here and I'll share things to that you can avoid if you're just starting your journey or if you're stalling out because of the same kind of things and how to move through them. It's a big partnership with CISO obviously for us, but I'll get into more specifics there. And then the theme on the data infrastructure and quality is really in a super cliche is garbage in garbage out, right? So whatever repositories you ultimately are exposing your developers to whatever LLMs you're connecting to, whether it's in house data sources or document repositories or whatever, there's obviously an effort that needs to go into cleaning some of that up beforehand. Otherwise to your point, like it's only going to accentuate or accelerate the drift, right? If you have bad information, I can get into some use cases there as well. Thanks for that, Kevin. So I think important to just take a step back, obviously fundamentals are important. So I think these are foundational elements that I think will contribute into a lot of what you're seeing as far as the pace of innovation at Axos. Quickly, we got another poll. Again, we're trying to keep you engaged. We know you're likely multitasking, but this is our effort to keep you engaged and active in the conversation. We're going to get to the meat of it. So second poll, what is your biggest barrier? If you say all of the above, you won't be alone. So please, if you can respond, helps us figure out sort of where to also steer the conversation. And we'll give it about 30. Some good engagement. Thank you for hanging on. Alright. So looks like top two legacy system constraints, ability to integrate and then governance risk and management with actually also the lack of clear use cases. So that's great. So I think the agenda we have is we'll hit on a lot of those three elements and what we discussed. So thank you again for the responses. So without further ado, really again, super happy to be a part of this session with Kevin. I think what you'll find is pretty fascinating thought leader as far as in the space of driving innovation at scale with low code. And I would say, again, low code, all of it is not created equal. I think we view it as enterprise AI powered low code application development. And so Kevin's going to talk a little bit about the journey that AXIS went to kind of really, I would say first start with, was it five years ago, Kevin, roughly give or take of the foundation that they've created? And then we'll segue quickly into a handful of pretty high impact use cases that they've deployed. So Kevin, I'll turn it over to you. Great. Thanks, Liz. So yeah, I started five years ago and give you a little bit more of the history of Axos and what I inherited when I got here and why they brought me in was was one of several tools in the belt, if you will, here. And none of them were being used extensively. OutSystems was really not a focus. There was no formal training in place. And so the team I had was only eight developers and they were supporting kind of this mixed bag, like I said, of tech, which included like K two five from a workflow perspective, SharePoint on prem workflows, Nintex. We also did some RPA and OCR tech, and then we had OutSystems. Right? And then there was a couple of apps that had been published, but they had been built by a third party partner. So the team itself wasn't really, hadn't really adopted. And I had to overcome this mindset, right? Which was this is beneath me or I have bigger aspirations. I'm a high code developer. I want to do C sharp. I'm dot net, blah, blah, blah. And the reality was even then I recognized that, and this is obviously before AI was really into the scene at all. It was low code, right? And so objectively assessing the tools that we had, I quickly came to the conclusion that OutSystems was the future. And so we heavily shifted to that. So we eliminated k two. We moved off of SharePoint for a majority of workflows. We've got and at that point, like I said, two apps. To date, we have 32 apps in production. We'll have 40 by the end our fiscal year, which is in June. So huge acceleration. Right? Part of that though was how do I convince the developers? And I did have some that left, and that's fine. And oddly enough, a couple of them that left came back. So they boomeranged back realizing that I was right, that low code was out there and it's the future. Part of it was, again, going back to giving them challenging work. So for those developers that were used to solving more complex problems, the work is there, right? So for me, it was more technical. How do I establish, you know, better integrations with our other cores, getting them on the API side of things, getting them additionally, getting focus on parity with the UI UX side of things with our other applications at the bank, right? So one of the things that unfortunately I came into owning was the OutSystems applications that had been developed were not using any kind of an Axos theme or style guide. And so the perception from business partners is that it looks weird. It's not what we want it to look like. It's got a canned UI, all these other things that I had to overcome. So how did I overcome some of that? I approached the business partners and basically said, what are some things that you could use some help with that you know are never going to get approved or assessed from a steer co perspective, if you will? Where do you have very manual or paper based processes that we could help you digitize? And so I partnered with, I just picked one business partner. He gave me a use case. We knocked it out. And then from there it just snowballed, right? Because now I've got, I had an advocate. He starts telling his peers about, Hey, you should check out what Kevin's doing on the OutSystems front. And that's how we won over the business. Right? When it came to additional adoption, so that was, you know, the business champions piece. This entry level training that I alluded to, there's nothing, there's no secret sauce here, right? We hire, we recruit college grads and we typically target computer science or at least finance majors, depending on the specific team they're going to go on. But we're just using the OutSystems Guided Paths. And then in addition to that, we're slowly introducing them to the team through either bug bug or defect fixes or small enhancements on existing applications. Right? It's the fastest way to learn. Now, when I say slowly ramp up, that's like three weeks. So within two to three weeks, developers are developing for us. Right? Within two to three months, they're typically on net new projects. Right? So it's pretty fast learning curve in my opinion. And it got to the point where I saw how quickly that this was being done that I felt like anyone could do it. And I told this story at the one conferences both in Lisbon and Miami. So if you were there or saw those, this is a repeat for you. I had a my son's friend approached me just over a year ago and said, hey. I'm looking for a role. I just graduated, and I saw you guys have this developer program. I said, yeah. What's your what's your degree? And it was a marketing degree. And I said, okay. Well, that might be a little bit challenging, but here's what I want you to do. Go do these guided paths. They're free online and do it before you get your interview. So he shows up to the interview two weeks later. He's gone through the guided paths. Additionally, he actually passed the cert for a developer, which was pretty impressive, which means he was obviously doing some things on his own and playing around with the platform. He's been with me a year and he's folded right in. He obviously had a little bit of a steeper learning curve when it came to like SQL, like he needed to learn some SQL and he needed to learn just some other things. But the point was that it's that easy to learn that as long as you find the right people with the right aptitude and the desire, anyone can learn this thing. So that was a big part of it. So this number here, you can see at the bottom what this 40 plus 125 is. And this really goes back to, I think it was Kevin Lang, think was the name Kevin asked the question, how are we using OutSystems across the enterprise, right? With the other lines of business. That's what the 125 developers are here. So I have 40 COE OutSystems enterprise developers or developers in my other development teams that are exclusively now using OutSystems. And just to give you a little bit more context of my consumer development organization, I have approximately 300 people in my org supporting multiple verticals within the consumer bank. That's also a mixed bag of Angular, Xamarin, Maui, C developers. Right? So it's all these different skill sets. We still have those legacy applications that need support, which is where a bulk of my developers are, at least for now. So the 40 that I have, and I'll get into it a little bit here, is how I use those in this distributed development environment. But these additional 125, and actually I think it's even more than that now, these are developers in these other lines of business, with our clearing, with our advisor services group, in the commercial bank, all these different areas where they've started to learn out systems, they're starting to build applications within each of their respective business areas. So it is widely used and that is a mix of use cases. And I'll speak more about at a higher level, what are some of the use cases here on the right. So Kevin, just to jump in to compliment a bit of the points in the first two before you get into more of the proving the value. I think what Kevin laid out is very consistent with what I've seen as far as adoption, not just the platform, but I would say getting the force multiplier aspect of burning out a backlog, as Kevin mentioned, 30 plus applications. Typically what we've seen is you start with a medium to kind of relatively meaningful application within a business line, right? So that whole partnership that Kevin was touching upon to figure out what are the demands of the business sounds pretty straightforward. But what we see is typically it starts within a particular unit. So that business and IT alliance there, and then that within itself serves as a foundation. And I think that the drawing down of the backlog is quite interesting to me because what oftentimes is overlooked that Kevin can touch upon is the acceleration can largely not only, but be attributed to the reusability of what you built. And so the reuse aspect of this, this is not spinning up a project necessarily net new every time. There's a lot of reuse from what you build from application A to B to C and onward. So some of that's important and it does tie into the value that Kevin will touch upon at Axos, which is we have to remember many players can spin up that initial application from kind of the design to the deploy, and we'll talk about this with AI, but the manageability of that becomes increasingly more important when you think about OpEx, right? So, hey, you can spin up a bunch of applications, but how do you manage that and how you start to manage a portfolio that's now spanning across multiple lines within the bank here in this example, and ensuring governance security and overall cost effectiveness and keeping that technology dead at bay. Right. And so we'll talk a little bit about that, but I just wanted, Kevin, add to that point you made earlier, which is I think the reuse starts to now, not only just with the learning curve, but that reuse element becomes crucial as you start to address other applications in the backlog. Yeah. And to that point, actually, when it comes to our internal apps, like solving internal workflows, the average dev time of an OutSystems app in that specific kind of use category, if you will, is only like ten weeks. And that's without AI. So historically, was one to two developers picking up a business case and basically modernizing either some legacy manual process or hybrid manual with, you know, some, you know, spreadsheets or some rudimentary SharePoint flow or something like that. So it is it is highly, highly effective. And we're at the point now where when the business unit comes to us and says, hey, we've got this process and we want to do this. And we have them just walk us through the existing process immediately my mind and the manager who runs the team, we are immediately equating that to an existing app we've already built saying, hey, let me show you this app and I'm going to show you how we did it here. And is this what you're talking about? Would you benefit from this? Because we have everything from very linear kind of workflows and out systems to very sophisticated recursive workflows, multi level approvers, approval matrices, all these things. And so to Luis's point, we've got kind of this catalog of apps and modules that are easily copied and put into new applications. Or in some cases, we've built an app where it's solving for a very particular business unit problem, but that business, that problem persists in other business units. And we've been able to just layer in those other business units within the existing app, but then we completely bifurcate the user experience and segregate what they see and what they're exposed to based on their business unit. So yeah, to that point exactly. Yeah. You got another point, Louis or? No, no, it's good. Was, I think you and I can talk all day about this, but I think people are looking at, okay, where are we, whenever you're to get to an agentic, but just a, I guess a couple of points on the value piece and then we'll push forward. Yeah. Real quick. The high code versus low code for me, it was if if the cost of upgrading is equivalent to just redoing it in OutSystems, we're doing it in OutSystems. Right? So that it started with a lot of end of life tech, Angular upgrades, Xamarin or MAUI upgrades, things like that, where we just decided to move it. And so that's how we ended up with a lot of things being deprecated from a legacy perspective. And then with that comes the benefit of not having to do patching and all the other back, you know, the infrastructure type of stuff as well. And then you saw those use cases real quick, but those use cases, you know, that was kind of a chronological sequence of, or an evolutionary thing where we started with just UI only type things, got into full stack and out systems. We did some black box service applications, meaning there's no UI. It's just providing a service between two other systems. We're now using it as an API orchestration layer. And then obviously now we're into the agentic space. Fantastic. And I think as you touch on this evolution with the governance, as far as kind of going back to some of the key focus areas from the audience on the poll of how do I get started? My concerns with governance, legacy, you know, interoperability. I think that's the one thing that, you know, as you work through this that I've observed within a lot of client and partner conversations is, you know, we say full stack development on our end. What does that mean? It's the composable architecture that lends itself, I think, to be interoperable with your existing tech stack. So you mentioned a lot of these other systems. What we refer to is it's more of, it's not just build or buy. I think it's build and buy. And so we often look at, I think, large enterprises that tap into ALKSystems as an orchestration or really that sort of collective kind of customer hub to support that integration across many of the core existing systems and different data sources with the ability to, quickly integrate, leveraging a very extensive and robust repository of pre built components. But with that front end experience, right? So we think about the full stack is obviously tapping into the data, the business logic and then the front end experience. But that interoperability I think is a key thing as you look and I, I'll, you know, point to the audience on this one is, as you look at evaluating different types of platforms and considerations naturally is what are the platforms that can actually bring in some of not just what you have in legacy, but the governance aspect. And then now you've got to fold in all of these different elements of agents, and how those plug and play and orchestrate effectively with one another. But over to you, Kevin, talk a little bit about the Axos journey and kind of your evolution moving into the AgenTic future. Yeah. And I think we'll get more into some of that. I see some more good questions coming up. I think I'll naturally touch on some of it, but we'll make sure we come back to them as well. But this whole slide is really less focused about the specifics of OutSystems and using OutSystems from an AI perspective. Because feeding on what Luis just said, we all have legacy or we all have other solutions in place that doesn't go away. It's a hybrid model. I've got heavy, heavy on prem in house built solutions as well as cores. And so I have to acknowledge that I need AI improvements outside of just what OutSystems can offer on those conventional or high code stacks. And so this is really speaking to just AI adoption in general, whether you're looking at it in OutSystems or not, and this will speak to the governance side. So the first thing for us was I wanted to make sure we partnered with our CISO because you've got to make sure you're not allowing some of these LLMs that are out there into your world, right? So not all of them are equal. Some of them are stuff out of China and stuff. You don't know. CISO involvement is a big piece and you get approval there. The other thing is you want to establish some basic usage rules for business and tech. What's appropriate use? What are the kind of things you do want to put out there and feed into it? Because you, know, Copilot, you can upload documents and so you want to box that in for sure. Part of it is, I would recommend that anyone who's, when you're getting into this, have a committee of some sort, whether it be business focused, whether it be tech, where potential use cases are being brought to the group, to a larger committee to say, does this make sense? The other side of this is that you have to, part of this paradigm shift and part of the mindset is people are going to be apprehensive to try new things because if they fail and it doesn't work, they don't want to be held accountable. Right? Or how is this going to impact my career? And so by having this centralized task force in our case or committee, you're bringing these use cases and then collectively you're saying, hey, we've got all these problems that we think we can solve with AI. And whether that be building agents or whatever the case may be. And collectively, if you agree and prioritize what those are, now you have buy in and people are going to embrace that experimentation and exploration without fear of repercussion, if you will, for lack of a better term. So you're going to foster that environment of experimentation, which I think I do really well. I have developers at every level coming to me almost on a weekly basis saying, Hey, I have this idea. What do you think? And it's like, okay, At my level, I can just say, yeah, I think that's a good use case. It's within the the guardrails, if you will. Let me know how it works out. If it's if it works now, how do I disseminate that and get it out there? If it didn't, cool. Move on to the next thing. No harm. It's Kevin. It's interesting. I know we're going to be getting to the use cases just right after this, but, I've been also seeing quite a bit of what has been referred to as councils that have representation from different parties that are, to your point, supporting where the investments go in the strategy and knowing that you could fail, you could improve, or it may not yield as far as expectations that are set as sort of that support infrastructure outside of just the IT element here from a business perspective that helps, I think, promote, motivate that innovation to move forward. Right. So I think interesting point that I've also seen and observed across many of my conversations. Yeah. And the other thing is by having that kind of centralized task force, if you will, or that governance committee or whatever, you're able to more effectively agree on how to ensure you don't have redundancy across teams. We have a very large org and focus efforts, but also when something is proven out, how do you ensure that it is being propagated and that people have a venue to come back? So outside of that, have like open AI office hours, right? With the leads that are in charge of a lot of this R and D or in charge of maintaining the usability of the workflows that we have. And I'll get more into those as well. So these three, this is kind of like how I view the evolution of AI adoption. And the first is basic. And I'll just tell you right out of the gate. This is just skip this. This is where we wasted too much time. And this was basically like, okay, we have AI everybody go use it. And it was very little direction. It was like, okay, we've got, we've got Windsurf, we've got Copilot, we've got our cursor and we want you guys to go use them. Well, the problem is that you give it to in my world, 300 people, I'm going to get 600 different results. Right? So it's where in this example, it's where users like in my case, like take a developer, a developer is just coming up with how to prompt it. There's inconsistency. I have developers across nine or 10 countries, obviously English as a second language across the board. And so there's inconsistency with how to prompt it. The results are going to be varied as well. And it's a little bit harder to measure. It's also where they're taking the work they've done and asking AI to fix it. And I'm going get to that, but that's, we need to, you need to inverse that thinking. We were using it for code fixing, right? This essentially is what it was. This is also where when you talk about AI from a customer perspective, you're just doing conversational AI for just maybe FAQs. You've exposed it to some repos and it's not doing a lot of actual thinking or insights. That you can actually skip on it, right? You don't actually have to go through those learning pains like we did, because if you move just to the intermediate, the focus here is, and this gets into one of the questions that was asked and a lot of the thematic here from people is how do I find my use case? I don't know what to go after. So what we did was in Myspace specifically, I took a lead or manager in each of the kind of functional areas and said, hey, QA, development, product, whatever. I want you to identify what your opportunities for improvement are in your space specifically. And then let's figure out an agent that can help with that or how to implement AI. And so that's the first thing is that's a good way to elicit that is just assign it to a small group within that area or an individual to say, what's a major problem? What's something that has an opportunity for improvement? Right? And I'll get into some more specifics with that. But this is also where once you get those use cases, the charge for that person or the team is, I want you to perfect those prompts, like figure out how to use Cursor effectively to solve your problem. And then once you've got it giving you results that are actually tangible, put that into a workflow. And then we can push it out to the masses and say, Hey, we've already built these workflows. You don't need to think about how to prompt it. All you need to do is feed it your work item. And it's gonna know this one's built for giving you code around this, knowing it already has access to all of the existing repositories, for example, right? This is also that paradigm or mindset shift going away from task oriented work and maybe focusing on feature level. In our world, we use ADO. So you're going to see a lot of ADO kind of hierarchy here, epics features, PBIs, those kinds of things. But what it means is have AI give you the context of the feature, not the individual task. The other part of it is have AI give you the initial code and have the developer focus on refining it to match the acceptance criteria from a business perspective, knowing they have that knowledge of how the business operates. And so it's an inverse thinking. And that's where you really start to see consistent and measurable results. This is also where the conversational AI, you're exposing it to document repositories, you're getting more insightful information. Again, goes back to the garbage in garbage out. So if you're using it for, say, a help desk assistant, you want to make sure that if you're exposing it to, for example, Salesforce cases, that you've gone through those Salesforce cases and refine the agent to say, here's what I want you to focus on, eliminate certain things so that when it gives the customer like a guidance on how to resolve an issue, it is accurate. This last bullet point here is also interesting. Used before we had the opportunity to shift a lot to out systems, which we're doing now, was end of life software upgrades. So we actually used AI to do Angular upgrade from very, very old Angular to current Angular. And we did that in 10% of the effort that was estimated by the team when we initially evaluated from just a traditional, we need to upgrade perspective. And we did that with somebody right out of college. He literally had never used Angular in our case. And he used Cursor and he upgraded our entire online banking platform in like four months by himself. So pretty impressive. Yeah. In the advanced state, the last one here, this is where you're getting into AgenTic Workbench, right? You're building agents for solving complex problems. And of course, everyone is looking at agents, AgenTic being, you know, kind of replacing human decision making or interacting directly with customers. And of course, we're looking at that. Most of my use cases to date have been heavily focused on internal improvement, I'll go into those. But this is also where if you've got the right insights, you're able to lead customers with, like I said, suggested actions or guided selling. So specifically in the financial industry space, this is where you've seen fintechs out there offering kind of canned solutions on this, but you can build them yourself. Right? And with the acceleration you get without systems, it's not daunting as it may have been traditionally. So specifically you look at things like, you know, you've got customer insights, they're sharing their spending habits, they're sharing other accounts. You can deduce certain things. You can deduce life events. Maybe you're going to suggest, you know, you see they don't have a five twenty nine. Don't have a five twenty nine, but you can tell they recently had a child or maybe based on their financial portfolio, you know, you want to ask them if they have a trust, you don't see that there, but they've got, they've amassed enough to a point. So there's certain things like insights like that, that you can help with. It's also like evaluating other products like, oh, we see you have a mortgage at bank X and it knows what our rates are and it can offer them a very hyper personalized offer to move their mortgage and save some money. Right? I also see this as where conversational AI starts to replace UI. So I should be able to come in and say, Hey, I want to move money. And I basically tell the AI, I want to move from A to B. Here's how much send it today. And it basically says, is this what you want to do? And you say yes and initiate. So that's where you're really starting to get into the really interactive conversational AI piece. So Yeah. Really And I think, as we segue now into some of the use cases, think, back to the, to the responses to the poll, polls earlier, you know, a lot of what you had mentioned before is how to be thinking about it. So whether you are just getting started, right. Kind of bypass the basis stage and start to really figure out where you're in that immediate phase to get the impact and get the alignment from the business and IT folks, and have that sort of counsel, right, or sort of a group or committee to support that effort. But what went into your thinking as far as starting with these use cases? I know you're going to talk a little bit, but if we can quickly kind of go through the logic behind these, you've talked a little bit about, hey, the actual software development life cycle and the application life cycle, as far as there's economies of scale there and inefficiencies that the platform already has, but you're augmenting that and naturally. And then there's internal areas and efficiencies that can be derived across the bank, and different business lines. And then naturally there's out to consumer, there's out to the external Right. But so can you talk a little bit about sort of why these and where you bought, you know, what helped you land here essentially? Yeah. And I view it as kind of, there's like three buckets, if you will, there's this backend or you know, the SDLC aspect of it, there's operational layer and then there's that customer layer, if you will. And my product teams are focused on the customer layer. Right. And so I don't need to focus so much on that, But there was a question I think that popped up around, you know, replacing human decision making. Yes, the answer is yes. Obviously that's being pursued in several capacities. I had a new app that came to me just in the last two weeks where the business said, hey, we've got all this business logic that's deciding, oh, this might resonate with some people. I wanna be able to move the logic when a customer requests a debit limit, a debit card limit increase. Right? And I wanna move it out of this legacy system and I want you to do it in out systems and we're gonna give you all the logic. And I said, I don't actually want your logic. I want the source of information. Right? What is the decisioning? Right? Give me their give me their spending history. Give me a history of overdraft. Give me all that. If I can expose that to an agent, the agent, the agent AI can make the decision and then you have guardrails, right? To say, okay, I'm gonna let the agent approve up to a 50% increase or whatever the case may be. Now, after that, I'm going to initiate a workflow that's actually going to go to human keeping human in the loop, right? So you can box it in, but those perfect examples of where, you you asked about how you can do It's in play. I don't focus on those per se, unless I they're just in my face. Right? They're coming to me with stuff. I focus more on these right here, which was my thematic here is how do I improve the SDLC focusing on every persona in the SDLC, right? And so first I started with, and we were an EAP partner with OutSystems on the AgenTik Workbench, obviously. And so one of the challenges we have is I have a mixed environments of ODC and OS 11, and I've got multiples of each. When it comes to the OutSystems 11 platform, for those of you who use it, and for those of you don't know, there's logs, right? There's enterprise logs in there. And those enterprise logs are just pages and pages of all your apps and everything going on. And most customers build some kind of a dashboard around that to figure out how to sift through it. I wanted an agent. So I built the agent to go into those logs in all of them. So it's an ODC agent dissecting all 11 logs and it's deciphering things. And we'll get more into that. Then the next piece was migration of legacy apps. And that was truly using an agent to reverse engineer an existing application. And then from there, that agent, we get these two other agents, which further assist in the process, whether it be with product BAs or technical BAs and then developers. So let's get into the meat of these. So the first one was the platform monitoring. So I'm going to go through this pretty quickly. It's a pretty simple agent that we built, but essentially what it does is you may have something that happens once and it's a blip and it may or may not be customer impactful, but if it happens a 100 times a day, I want to know about it and I want someone looking at it. So this agent is not only categorizing the criticality of errors, knowing that, you know, a 500 error is catastrophic and needs to be escalated immediately as opposed to some others where I'm not really worried about one off, but maybe I'm worried about it after it hits a 100. And so it's aggregating those frequency of occurrence and then depending on some thresholds, it's creating alerts, right? So it's essentially a monitoring system of 11, knowing that Mentor and ODC is essentially going to be doing this for us, but I needed something in eleven. So that was the first use case. Pretty simple. We did this in, we built this in like two or three days, I think. So if we want to jump to the next one. Yeah. I think is the next one that have a little video or no? This is the saying I got a request Actis. So let's see if that happens. If not, we'll probably push on then. So, so yeah, this was a, this is really showing the back end of legacy modernization. So let me just explain what this one's doing. We connected an agent to our Azure DevOps repositories, and we chose a use case in my in this one was, and I can speak specifically about this. There's nothing proprietary about it. We have this dot net app that our QA team uses for spoofing API data. So most of you probably have the same situation where you either don't have enough lower environments from vendors or vendors limit the amount of test data that you can have or the volume or whatever the case may be. So we have the ability to spoof. So the spoofing was a dot net app where developers were, okay, great. So now you can see it actually works. So this is the agent it's going through, it's pulling the code out, it's grouping it by features logically. And then it's going to create a discovery to our requirements document, right? So that requirements document is essentially extracting all that logic and saying, here's what this app is doing and here's what you want it to do when you build in and out systems. And what this app is, like I said, it's got this backend, but the QA team was manually creating Postman calls that would call these endpoints. So there was no UI to this application as it existed as a dot net app. We then took this, the output of this agent, which is this requirements document. And we fed it into the Mentor AI analysis systems and it produced the app, right? And it's, this one took us, I want to say two weeks, I think to build this. Obviously we had to work through the ADO integrations and all that, that was new to us. And then we tweaked the requirements document a little bit, but essentially what you're seeing here is a real time example of the user not not manually modifying anything, running the agent, he'll get the output that you'll see right here. So he's gonna grab this this text doc. The text doc is the requirements document. We've built a template inside of the agent to conform to what is conducive to Mentor AI. We drop it into Mentor AI here. And I did shorten up the processing time there a little bit, but all in all, it's only about eight minutes to run this whole thing, eight or ten. And then we got an app, but what's really interesting is that it created these screens for us, right? So it deduced that you need a UI around this and we're going give you all this stuff. Right? So we're just quickly going through these to show the output of this, but this is how usable, right? So our QA team can one off spoof data without having to create postman calls and do all this manual stuff. Right. And Kevin, we're now on the BA assist. So you want to talk a little bit about sort of moving right on the agent assist side? Yes. So what we did is I took the legacy modernization agent and said, you know what, what if I want to jump into that middle before I go to Mentor and I want to give the business an opportunity to review what the old app is doing before I'm going to modernize it. So this actually allows a BA, whether it's a technical BA on my team or a product BA to go in and say, Hey, what is this doing? Or I mean, how many of us get a question from the businesses, Hey, what is this feature doing in the app? And it's something that was written five years ago and nobody's here that knows what it does, or it was poorly documented. So if you, I think the next screen if we kept it. Yeah. So basically the way this works is like, you can see here, tell me about the business logic for the bank vertical. Okay, here you go. So what it's doing is it's going through those same code repos, but it's now putting it into business verbiage. It's not technical jargon. It's saying, here's what this does, right? More technically on the right, you can see, can you please explain to me how file attachment works in this particular feature? And then give me also the link. Yeah, here you go. Right? So it's twofold. It's one education for the business to understand how things work. But for us, it's also going back to the product team and saying, this is how this works. Is this how you want it to persist? Do you want to take an opportunity to modify any of this before we build the app? Okay. So that was the next evolution of this. So this is helping the BAs in that capacity. It's also helping product. So if you go to the last use case, which was the Dev Assist, so this was also, okay, I've got work that the devs need to do to refine things, right? And this is another short video, but what this agent is doing, dev can say, Hey, here's my work item. Tell me what I'm doing here. Right? What am I working on? Or how should I be thinking about developing this? And this agent is actually recognizing at what level of hierarchy the work item is that they've referenced inside of our Azure DevOps. So if they've given it a product backlog item, a PBI, it's going to give them context within that PBI, criteria, basically give them a summary of what it is they're trying to do. It will they can also prompt to say, tell me how it fits in to the larger feature or the epic. Right? So they can get a more comprehensive understanding of what it is they're building and how it fits into the larger picture. It also works the other way. So if they, if they give it a feature or an Epic, it will drill down into all the affiliated underlying artifacts or items and say, here's everything you should know. From there, they can also start the prompt say, Hey, what, what work has been done on this? Who's already worked on other PBIs in this feature so that they can collaborate. Where this is very powerful for us is that I have such a huge dev team. It's not feasible to do walkthroughs on every project kickoff with a 100 developers or across all the teams. And we focus more on like leads and AVPs and things like that. But now I have the ability for these individual devs to do their own phishing, right? And get the information they need and find out who they can go to for more information as well. So yeah, that was the final use case we had. And then getting into where we're going next. So if you wanna just jump, jump this one. Yeah, the QA assist agent. So that's where we're going to build an agent for QA. So this is going to, we've already got some of that going and it's basically giving them the test cases based on the acceptance criteria of PBIs in this case, as well as the next permutation of this, which the team brought it to me yesterday and we're going to start it probably today or tomorrow is unit tests. So at the dev level, using an agent to help ensure we have comprehensive unit tests before it even goes to QA. So I expect those to be turned around in the next week, maybe ten days. And then I also have teams that have come in where they built agents in other platforms and said, Hey, it's working, but I want to move it to the agent workbench and OutSystems because inherently that's where we're centralizing things, but they think they're going to get a benefit out of doing that all centralized. So we're moving to that as well. So I know that was really quick trying to be cautious. No, we appreciate it, Kevin. I know we had a little bit of technical difficulties on the first video, but thanks for pushing through. I think impressive nonetheless, far as, as you mentioned, right within the SDLC process, there's already things that you're doing. And actually those are going to extend and provide benefits as you look to use cases that are out to consumer and even internally across the bank, across different business lines. So we've got our second to last poll, so thanks for hanging with us. And just, yeah, if you could give us a sense of where you got, where you folks are prioritizing would be great just to get a pulse and then we'll follow-up with one more poll and then we'll wrap up here shortly. Looks like the core business ops is out in front, which is where like my bread and butter has been. Right? So when you talk about STP straight through processing, that's where we have OutSystems apps where initially it's been, you know, we built the business logic into that. So for example, we have an app that we built that assesses the risk of money movement and specifically wire transfers. And we've got over 90% straight through processing on wire transfers now without a human having to review it. And that spans multiple lines of business to the tune that that application alone has offset our headcount by approximately 30 people. Wow. That wasn't part of the script. Thanks for adding that Kevin. I think we're getting into one more. I think Kevin's got question. One more Kevin and we'll get into Q and A. I think we've got enough time, Aaron, you can be honest here, but the last quick poll and then we'll open it up for Q and A. Erin, how are we doing on time? We're good. We've got around twenty minutes to reserve for the Q and A portion. We'll wrap this poll up. Appreciate the 28 or so hanging on. We know it's a little bit longer than usual, but hopefully those have stuck on. We'll we'll get some additional value here in the Q and A piece before I wrap up. So great. Looks like, all right, call center optimization and knowledge discovery. Probably not surprising. But before we add any additional color to that, Kevin, you want to go ahead and come off mute or I guess Aaron, we can see if he can pose this question. Go ahead. Thank you. Yeah. Not confusing having two Kevins on the call. So Okay. I I was actually really curious about the off scripted answer that Kevin just gave about the the wire transfers aspect of it and and being able to reduce the headcount on that. In that particular scenario without, you know, giving away the the company secrets on that one, is it actually like an authorization transaction, or is it kind of a post monitoring type of thing? Did you allow AI to actually go in into the authorization flow? Yeah. Yeah. It's a good question. So we're actually, this is the approval process before the wire is executed. The risk mitigators, right? It's OFAC, it's all of your account verification, it's beneficiary validation, etcetera, etcetera. And so it's going through and assessing all those third party services you use, establishing a risk rating. And then it's also looking at a lot of internal things, right? There's thresholds around activity on the account, various things, right? Like have they recently changed you know, login credentials or contact information and all these other things? And it's doing all these things literally, think are, and it's, I want to say it's about a thirty second process to go through everything. But yeah, I mean, obviously I mentioned earlier, have heavy volume of wires, well, just across all the lines of business. And they're not all in this app yet, but we do a large number, large volume of wires on the consumer side through there. And that's a perfect example of, you know, you talk about core competencies or backend office type functions of our 32 apps, 27 of them or so are back office type of operational workflows. And if do you mind if I ask another question about maybe some of the the data underpinning that? Data underpinning which piece? So like actually the whole process. So you're talking about the different evaluation points as things are coming in. Right? So obviously that's a large dataset. So you have potentially like a warehouse or bricks or something behind the scenes on that. But did you do you have multiple models, like specialized models on whatever the particular platforms are that then you're consolidating it all within OutSystems agent to then query these different specialty sources? Yeah, yeah, absolutely. Because you have, the validation of a funding wire on a loan is very different than a consumer C2C wire, right? You're using different vendors for risk mitigation there. You're validating an escrow company and you may be using white lists, right? Particular companies. So yes, there's different, there's absolutely customized. And that was when I mentioned earlier, when we started with an app and then we found these other areas of the business that we've been able to layer in, that's a perfect example of one where we started with just consumer and we layered in these other wire types from other lines of business to achieve that. We've done the same with other internal workflow type of apps where, you know, audit type things like where we want to initiate internal audits or versus external audits and things like that. So, but again, most of those internal apps were solving for a very specific business process that again was just antiquated. The business had been asking for something for years potentially, but there was no bandwidth or it was cost prohibitive to do it in high code, right? Or it was, and that's how we got this environment of like sit dev and SharePoint. Well, what inevitably happens when you do sit dev, the person who built it leaves, nobody knows how it's built. Or what we were running into is it's reached the end of its useful life. You've reached the maximum capacity of a particular platform. For example, K2, we were on a pretty antiquated version of K2 where it did not, it was not conducive to recursive workflows and recursive for me being, it's going to go from A to B to C back to A then to C, right? And we have the ability to do that. I think there was also a question about approval. I don't remember when it was about approvers inside of workflows. Yeah, that was actually mine. Have you considered using it? Yeah. So yeah, we do. That's part of the opportunity is that a lot of the apps we built predated the agentic piece. And so now it's going back and retrofitting to say, Hey, I can replace that with an agent knowing that I'm going to expose the agent to those data sources and it's going to evolve and get better. Right? So in that case, in let's say that approval workflow phase. So then parlaying, I think George had a question about governance. So then do you create like a governance for that particular agent, let's say to access that operational data? Yeah. I mean, it's yes, you have to establish those within that, within each of the agents, right? Because if you don't intentionally box it in and tell it do not do certain things, it will ultimately do those things. Right? So I think a perfect use case of that is United Airlines. For those of you who don't know, United Airlines was one of the first companies to go out and use AgenTic. And what they did is they, their use case was we want to expose it to all the customer complaints, and then we want to have it propose policy changes. And so they gave it this world complaints and let it decide. And unfortunately no governance around it. And it actually published a revised, the number one complaint was, I can't get a refund on my ticket. So it published a new policy that was, oh, all tickets are refundable and it published and it went live. People bought tickets and then they were just canceling and United figured out like, oh, oops, we shouldn't have let it do that. There was a lawsuit they lost. Right. And that's exactly the point. Like you still need human in the loop and you also need those guardrails. So I mentioned earlier, like you've got, you know, the debit card example is a good one where, okay, there's reasonable things that it can deduce. Like if you see that the person's average daily balance is a $100,000 and by default, you've given them a thousand dollar withdrawal limit then clearly there's an opportunity to approve that, right? If they've never been overdrawn, if you see they have regular deposits, you've also got typically exposure to their other accounts outside of your own bank right now with open banking, open finance, right account aggregation. So you have a lot more insight so you can make those logical decisions. Now, if you have somebody that comes in and says, I want a $50,000 daily limit. Those are where the guardrails come in. You either set a percentage cap or whatever, but even then you're not denying it. You're just putting human back in the loop. And I think that's the important thing within each of those agents. You have to have the wherewithal and the foresight to say, what are the negative tests here? What are the negative scenarios that I need to be cautious of where this thing, if it goes on its own, we're going be in trouble. Right. So I think the answer, of course the answer is yes. In I response to your, the guardrails thing is at both of our industries is financial. So I think there's a lot of pre work in terms of the policies and procedures and lending guidelines that we have, that as long as those are taken into account then. But then when you drift off of those strongly defined lines, that's where your negative testing thinking, to your point, becomes crucial. Yeah. It's that and it's also agents upon agents. Agents serving agents, right? Where you can start to build an agent that its entire sole purpose is to evaluate the other agent, right? And look for that Drift, or you can even have that agent box things in. And that's where we haven't done anything like that yet, but that's absolutely where I want to be going is looking at, okay, what logically makes sense there instead of putting it inside of there an individual agent? Is there an opportunity to encompass multiple agents with a governance agent, if you will? Great. Thank you so much, Kevin. I appreciate the question and the follow-up. Thank you to both Kevins, I should say. So Erin, do we have any other questions from the audience? Had one from One of my biggest struggles is forecasting AOS by use case. That's challenging with apps alone. How are you forecasting consumption across apps and AI? Great question. So I actually do track this. So I have a spreadsheet I maintain that I probably should build an agent for, to be honest. But actually categorize right now, I know my average AO is across everything. But I can tell you that for me personally, if I if I get rid of, like, the large the larger, like, customer facing apps, things like that, some of the more sophisticated, like admin portals, HR type systems. My average my average AOS per app is somewhere in that 50 range. So all these internal workflows, I have some as little as nine AOS, and then some that go up from there. But on average, it's probably, 50 on average for me. And then I have apps that go up. Right? So when you get to external portals, I have apps that are 300 AOS. I'm expecting to launch an app next year that'll be close to a thousand AOS. Actually, have three apps next year that will be close to a thousand AOS. Those are very, very heavy hitting core type apps, know, hundreds, hundreds, hundreds of screens, things of that nature. So it is a challenge and you do have to kind of establish that history of building, but hopefully that helps you in some context. You know, directionally it may mean nothing to you, but I think overall, it's probably not a bad swag. Yeah. And just to quickly add to Kevin's point before we wrap up, I've worked with Kevin on part of this sort of t shirt sizing exercise. And for those in audience the that are not familiar with AIOs, AIOs are application objects that represent screens, tables, integrations to kind of a simplified sense. And so on average, we see a range per application of anywhere from 50 to upwards of 1,000, depending upon the complexity of that given application. So I think doing some T shirt sizing, working with AltSystems for those that are clients and others that are not, it's actually an option that seems to be, although sometimes a little bit tricky, more predictable than some of the user based pricing that's out there, right? That you're dependent upon. But Kevin, as far as before I wrap up, and maybe this is more to be determined, you've got the kind of the existing application object construct that OutSystems provides around pricing, and there's other add ons and things that you and I know. But, as far as on the AI side, like how are you thinking about managing costs there where does of OutSystems fit? But what's your thought process there? Yeah, I'll be honest, like the benefits I've and we didn't even we kind of glossed over, but the benefits I'm getting out of OutSystems just far outweigh the cost of doing things traditionally. So to be honest, while I am cost aware, I'm not slowing the growth. I want best world class solutions. And so hyper focused on the size of them. Knowing that when you going back to my assessment and when I have the business partners bring use cases to us and they say, hey, I can go buy this product off the shelf or we can build it. And what it always comes down to is, okay, maybe that your one cost is gonna be the same for me to build it versus you go buy it. But after that, the run rate is marginal. Right? Yeah. I think Gartner states like 80% of TCO is within OpEx. Right? So, yeah. And that's the other thing is that when you talk about agility, that's we get that's what we get with OutSystems. I mean, my OutSystems developers, and I didn't get into the specifics, but we did kind of like these high code competitions, if you will. I basically took two dev teams, HighCode and OutSystems, put them on the same project and said, I'll eat the cost. I want to see who gets done first. Right? And every single time I did that, the OutSystems team ran circles around the conventional teams to the tune of it took them 40% of the effort. So, and the bug rate was a fifth. So when you look at cost of ownership, like, yeah, you could focus on the AOs, but I guess it's relative to what you have today and what it's costing you. And then you factor in, for me specifically, the whole other aspect of having this the platform construct, meaning this fully integrated platform from a DevOps perspective. And I'm not now burdened with patching upgrades, you know, all the things that go with what we do I specifically have to worry about with on prem as well. So to me, it's like I said, the benefit is far outweighing what I'm the alternative fuel. Yeah, well, thank you so much, Kevin. I know we just got a couple of minutes left. Really appreciate your insights and great questions. I wish we could field some more, but happy. I know Aaron will capture some of those and we'll follow-up, but to wrap up in the next couple of minutes, we covered a lot of ground. And so hopefully the conversation helped provide some insights and hopefully some nuggets that you can take away around how scaling AgenTeq AI from pilots into production is possible in financial services per the Axos story. And again, OutSystems, we feel, an option that you can leverage to tap into your existing investments, right, and harnessing the power of AI To, as Kevin mentioned, really focusing on not just building those custom apps more quickly, more cost effectively, but leveraging AI agents along the way to streamline operations, as we discussed. Overall experiences, be it internal or external, and then naturally helping you grow your business and wallet share. And then part of this is also just empowering the existing team to deliver faster, more effectively over time, future proofing your solutions. So as we know, there is a lot of different options you have. Hi Co plus Co Pilots, Vibe Coding, that some people think will carry into the future. But still many organizations, we see this, and I think Kevin touched upon it, are struggling from moving from pilot to production. So what is production ready? So as you think about those evaluations or those opportunities you have to tap into your existing, but also bring technology that can help not just generate the application, but help orchestrate and manage that over time cost effectively, OutSystems is an option that we think you should consider. So I'll punch kind of through, I'll kind of bypass a couple of things. But what I would say I'd leave you with here, relative to agent workbench, which Kevin touched on briefly, I urge those in the audience that, and again, thank you for staying on to learn more about that, right? Whether it's in the form of going into OutSystems website, you can look at the link here and just start to get your hands on the keyboard and have some of your folks work with it. And then, we also have other assets that you can see as far as what that looks like and what that could mean for your development teams. But again, I think you saw in Kevin's kind of journey where it started with four or five and now scaling up to 125. We see different sizes of organizations tapping in to the platform and supporting a pretty significant scale of application development on the platform. So it could be teams of three generating many applications to teams of 100 that are generating hundreds of applications, right? So it really runs the full gamut. But with that, I wanted to thank everyone for attending the session. Hopefully you got meaningful information and insights. And thanks again, Kevin. Always a pleasure to have your thought leadership in this space and your partnership for the last five years. And thanks for when you stepped into Axos that you actually took the time to evaluate OutSystems and now have been on the journey and realized value with it. So we appreciate your partnership. Yeah, with that, I think I'll leave it. Erin, I think anything else in closing? Nope. Just a reminder, we'll follow-up with the recording and slides within a week. So stay tuned on that, and we'll make sure to follow-up directly with any questions we weren't able to address within the Q and A. So thank you all for your participation, and have a great week. Thank you all. Take care. Thanks.