 Hello everyone, welcome back to day two of theCUBE's live coverage of SaaS Innovate 2024. I'm John Furrier, host of theCUBE with Dave Vellante here, instructing the citizens, there's a lot of AI discussions. Real use cases put into production, a lot of first news announcements out here in the industry, SaaS is selling lightweight models, there's big news there, a variety of other solutions. You're seeing a lot of enterprise adoption starting to come in, we've got two great guests here, Stu Bradley, senior VP, risk fraud with SaaS, assistant VP of fraud analytics and strategy at Credit One Bank. Guys, welcome to theCUBE, here at day two, kicking off. Thanks for coming on. Thanks for having us, thanks for being here. So there's a couple of things that jump out of me in the keynotes and all the news. One was the saved prompts, it's weird, but I like that little feature because it shows where it's going. Promptless is in our future for the language models. But the other one is the models. One of them is fraud detection right in your wheelhouse. You start to see the models, lightweight models, available almost like Lego blocks to be added in for speed. So I love that announcement. This is going to be a real opportunity for developers in your area, Stu. Tell us about what's going on in the risk side of it in your world. Yes, so look, I'm very excited about the announcement. The whole goal behind this is to meet our customers where they're at from a technology and an adoption standpoint. So being able to provide an architecture to extend across a multitude of different use cases, but also to get more advanced and provide models for things like solvency risk of financial institutions for fraud detection, which we've been providing models for almost two decades now. And using those as templates to ensure that we can further expand the horizons on the types of models we're providing to serve not only financial services companies, but organizations across different industries. The long, the chasing the tail of fraud has been going on for decades. Absolutely. This is a big day-to-day thing. We started theCUBE 14 years ago. Duke was supposed to be the solution. It was actually mentioned that the keynote was kind of funny that it was brought up, but now it's getting better. AI is going to be a solution. Absolutely. What's the current state of the art in your mind at your bank on how you guys are attacking the fraud? We saw the results on stage today, better performance on approvals on things, operational efficiencies, but fraud's the big target problem. That will take us through your perspective. So when I think about gen AI, there's a couple of things that come to mind. So the first thing is you have LOM models, so such as your chat GBT, but the other is your deep fakes, right? So these are really financial concerns, whether it's from authentication perspective or verification perspective. So in regards to chat GBT, I'm not sure about you guys. I mean like two years ago, three years ago, I get scams, email related emails, and they're coming back with a lot of grammatical errors, a lot of vocabulary misspellings, and a lot of which doesn't really necessarily make much sense at all. So there has been an improvement in terms of fraud communication styles that has been coming in. And what that ultimately translates to is a lot of more sophisticated and enhanced scams situation that we have been seeing lately, as far as the past year and a half or so. So these are your men in the middle type of attacks, your romance scams, your employment scams, and then to shift a little bit as well. So the other things that we've been seeing is the utilization of deep fakes, particularly as far as enhancement and using your, trying to bypass your, let's say, your voice, your voice biometrics, and then also your facial recognition applications. So all of that translated into risk that we're looking as far as how do we stop it, how do we prevent it? So last year was kind of an interesting one in the banking world. You know, normally when interest rates go up, it's good for banks, but then you had that situation last year where folks are having to mark the market, and then you had the first republic, or actually SVB first, and then even FTX, which actually had a pretty viable business if they weren't like skimming. And so what's happening now in the banking community? It seems like things have stabilized, but from your perspective, Stu and Long, what are you seeing today? So look, I think that was, you know, became on everybody's radar, just based upon some of those crises that we saw with the financial institutions you've mentioned. I think part of the issue is that many organizations have grown their risk management capabilities in a bunch of silos. Some were specifically geared towards reporting to the regulators, others geared towards looking at credit loss, stress testing their environments, but nothing is really forcing them to look holistically across their balance sheet. And so I think as we emerge out from that environment, the capabilities of integrating data, so you can get a holistic view across the balance sheet, stress test that, change different micro and macro factors to understand the impacts is going to be a really important response for financial institutions to think about. So when you say stress test, give me the what ifs, and you guys can help model that, you're saying from a holistic standpoint. What's your liquidity going to look like, if interest rates fluctuate, and mark the market versus. And keep in mind, I mean, this is the first inflationary interest rate environment we've had for 25 years. And so the risk managers today aren't used to in rising interest rates. So that's not something that they're accustomed to. So absolutely, I think part of the issue with technologies that have been available in the marketplace today is they run a very limited number of different scenarios against which you can stress test the balance sheet, liquidity risk of the bank for example. And the ability to run a multitude of different simulations is going to give a broader view and broader perspective that would allow them to do a better job of managing the overall risk of the bank. So just to follow up for you, Lon, I mean, again, you think for a credit card company, higher rates would be better, but then again, there could be more defaults. How do you think about the current environment and sort of modeling your business? So I think, so I'm going to quote the CEO of Current One Bank. So I think he really poses very well. So he said that it's better to leave a dollar on the table than to incur a $2 incremental losses. And I think that really summarizes pretty much the landscape that we're currently in. But as far as overall impact, I think the biggest impact, kind of taking this back a step as well, is be able to ability to kind of go across different departments and really communicate as one. Because what we have is to just do this point is that it's silos, right? So from a financial perspective or financial net fraud or net loss perspective is that there's a multitude of different variables that come into play, right? So you have, you have recoveries and then you have the gross fraud that we can control. But also you have other variables such as acquisition risk, right? So throughout the whole entire life cycle of the portfolio, there's very different perceptive perspective on what sort of mitigation strategies you'll be able to adopt. And I think by combining all into one, it really be able to help facilitate those changes. A question on the analytics side, because this is when, again, one of those things where you have now more data points. The more data you have, the better the AI is. We've seen that and that's been better. Performers is coming on, you guys are doing, addressing the performance challenges that are needed, more GPUs are available. What data points are now available or what AI aspects are in play that can help take advantage of these new data points? Specifically, like if I'm in Vegas, my phones tell me I'm in Vegas so obviously I'm not in California. So these are opportunities to get data made that up but that's kind of an example. What new data do you see on your side that JNAI could take advantage of today that wasn't there, let's just say a few years ago to help prevent the attacks? So from the credit cards side of things, is that recently we have migrated over to a new fraud detection engine strategy application. And what it allows us to do is be able to create additional custom profiles, whether it's based on the merchant level or whether it's based on the personnel or our customer's level. I think at the end of the day, it really translates into know your customers, right? The more you're able to build upon that and then the better the strategy and then the lower your strategy the clients would entail. So that's kind of really a win-win situation. Higher revenue, lower the clients. And Stu, what's your take on this because is it operational efficiency? Is it driving more revenue or cost recovery? Take your thoughts on this. So as financial institutions have been in an arms race against each other for better digital engagement with their customers and enhancing the overall customer experience, it's created, the attack surface for fraudsters has been expanded and it's created that additional attack surface which means you need to have additional data points. And we refer to it as creating a data ecosystem, not only of your internal data, but your external data, public records information. How do we validate an identity and you are who you say you are, physical information, biometric type of information that was mentioned previously. Is the device that you're using of high reputation or has it been engaged in a fraud attack with another institution? So being able to leverage different sources across the ecosystem at the right time in your decisioning cycle is tremendously important. So it's about creating that data ecosystem first and then the artificial intelligence comes into play now that you have those new data sources and it allows you to do a better job of modeling for new trends that you might see from a fraud attack. And what's different now in Europe that you're seeing? What's different now with GNI? What's GNI enabling from a disruption standpoint in a good way? That's different than it was a few years ago. Is it more data? Is it more compute? More workflow, augmentation? What is the key thing that you're seeing now that's in play? So aside from what Long mentioned previously, which is it's enabling the fraudsters to have better attacks. Yeah, exactly. One of the things that we saw, we just released our biannual association of certified fraud examiners state of fraud reports. And what's really interesting is that when you look at the reports we did in 2019 and 2022, in each of those 25 to 30% of financial or of organizations, specifically financial institutions said that they were going to adopt artificial intelligence within the next two years. When we resurveyed, only 5% have actually gone on and adopted artificial intelligence in a production manner. And it highlights the overall challenges of doing it. Challenges of getting the right data. Challenges of understanding are there any biases in that data? What are the downstream consequences of the decisions that you're looking to make? And so based upon that, we've seen financial institutions adopt it more to drive efficacy in their operational processes. How do we take mundane human tasks and automate those and make it more efficient but a more rewarding experience for those employees? The example I get, take a fraud investigation. 80% of an investigator's time is spent collecting information, pulling it together and synthesizing that. Well, we can use Gen AI to do all of that for him and allow him to spend 80% of the time investigating suspicious activities such that we can actually address the fraud problem. So George Kurtz, who's the CEO of CrowdStrike, he has this phrase that he coined years ago is mantra for the companies, stop the breach. And at the time people said, you can't stop the breach, it's just a matter of if, not when, or when, not if. And he said, I don't accept that. So I think you are like an analog to that, stop the fraud or stop the theft. It's hard to do that, but when you think about the compression in time, it's all about speed. Think about 10 years ago or even more, like you used to have to go back through your statements and find the fraud, right? Now we're checking out and it's like, yeah, it is me. Okay, so what's the technology behind that that has enabled us to compress that time so dramatically into near real time? And when you talk, question first part, second part is when you talk about AI adoption, you're talking about kind of legacy AI ML, and if that's a proper term or gen AI or both, I wonder if you could add some color to that. So the timeframe in which you make decisions is tremendously important. And once you have allowed a payment to go through, the recovery rates on identified fraud after a payment has gone through are in this one, two, three percent range. So the money, once it's gone, it's gone. So you need to go from a post payment and trying to chase the money into a prepayment analysis. And so we've been in that space where we're running AI based models within 30 milliseconds so we don't impact the customer experience at the point of sale for a card transaction on a digital card, not present transaction, or any digital peer-to-peer payment that is out there. And so that type of technology has been available. And to your point, I think that's part of the reason why AI hasn't been widely adopted because not only do you need to make the artificial intelligence give you the right results, but you've got to make it perform in the matter of 30 milliseconds to be able to stop the payment before it happens. And I know in the card space, they're all over that. But you've dramatically, as well, reduced the false positives. Absolutely. Yeah, false positives, right? Which is kind of annoying for the consumer, but those are way, way down relative to five, even five years ago. Right, so as far as technology, it's really a double-edged sword, right? So the faster a merchant or such can be breached, it also enables us to pick up those trends as well. When you have a velocity of, let's say, 100,000 transactions being attempted in a very short span of time, it goes both ways. If I'm seeing that in real time, which we have, I guess we're blessed to have real-time data at this point, is that these are sort of really suspicious type of activities and we'll be able to kind of really stop these really at the gate, right? We might be able to let a couple through at the very beginning, but overall, I would say the bulk of it will be declined. From a recovery perspective, what Stu mentioned is that it really depends on what the type of transaction type you're facilitating. If it's Karnaw present, maybe from a breach perspective, we should be able to recover some of those. Car present? So you're saying, okay, maybe the first, maybe one or two, whatever, a small percentage gets through, but then what, you identify the pattern and you're able to match that and you can do that without AI is what you're saying, right? So, because these customer profiles are built to tally up or aggregate the number of transactions in a very short period of time, a lot of what you could really specify in terms of what you're looking to get. Do you want to collect the number of transactions, 20 minutes, 10 minutes, three seconds? All those are possibilities. So we try to utilize those type of technologies as far as fighting fraud. And your point about the attacks are so much better. I mean, those phishing emails, those, it sounds like beautifully written, you know, and if they get personal information from you, public profile, social media, really, that's a game changer on their side. So how are you guys catching up? Because again, the attackers, they can throw everything at the wall. They don't have compliance to deal with, right? So they're faster, but we're getting faster too on the good side. What are you guys seeing on the speed to meet that threat? How would you scope that if I asked you that question? So I would say about a year and a half ago, we have seen an increase in synthetic fraud types. So your synthetic identity, so that, and this is quite common throughout the whole entire industry. So a couple of things is that fraud after continuing to be on the rise and as far as the current horizon, and then also you have kind of your first party fraud as well. So all these, as far as mitigation strategies, usually we kind of work with our partners as far as be able to identify these sort of risks or appetite at the very beginning during the acquisition side. It's almost a fun job at the weird, well, because it's almost like gamifications, like us against them and there's that cyber action too. It's cyber money. I mean, everyone wants the cheese as they say. They want the money. That's a big part of what you guys are vulnerable. I think other than the deep fake that I was joking about the cube and making sure we don't get deep faked from this conversation, I think one of the most frequent conversations that I'm having or questions I get asked from banking executives is what should we be worried about next? What scam is on the horizon? And when you think about it, that's actually the wrong question to be asking because the scams are gonna be consistently changing. They're gonna be consistently improving. The question they should be asking is what do I need to do to be more agile in my business to be able to respond to whatever is next? And I think the way organizations have grown in solving a problem with a specific technology with cumbersome integration is a big technical challenge and rationalizing that down to a less complex infrastructure to make more holistic decisions across the customer life cycle is absolutely the path forward to get that agility. To your point, it's a data problem too, right? All data is available. Your location, your history, everything's at the table now. That's going to be, I think, the opportunity. So you're saying simplify, eliminate the data stove pipes. Yes. Okay, build a holistic model of your organization and then conduct analysis in near real time. And that's kind of the starting point, I guess. And then it's iterate and get better and better and better. Laying down the infrastructure to be able to do that and then understand where you're at on the AI maturity curve, right? And we all talk about, hey, we need generative AI. But go on a journey. Understand what types and variations of analytics and artificial intelligence are going to be used to solve the specific problems and deploy those technologies and work your way up that maturity curve over time. Does that necessitate a change in sort of skill sets? I mean, I think about pre-cloud, you had like somebody provisioning storage loans and somebody else provisioning servers and the network person. Now you got a DevOps person who does it all. Is there a similar analogy in your world where when you consolidate and you simplify, you now have a chef that can do multiple tasks? Well, I think, and I'd love to get Long's opinion on this as well, but you need to take a perspective based upon what use cases you need to solve and what skill sets you have. So going back to that common architecture with a governance structure to be able to achieve that, and there's opportunities to work with the partners. So customers, banks, other organizations should focus on where they can, their core competencies, where they can add differentiation and look to lean on others where we can create models and serve those to organizations as a service when a night might not be part of their skill set or core competency. Student law, thanks for coming on theCUBE. Congratulations on your success. Final question to wrap us up. What's the most exciting thing you're working on that highlights the value of AI that people who aren't in the industry might not know what AI really is? An example of some of the things you're working on that you're excited about, we'll start with you. Wow, some of the things that we're working on, I think it has to do with a lot within the dispute side of the house because I'm mainly from fraud and I think customer experience is very, very critical. You have recently been either breached and you want that customer experience and then you want to be able to provide customers that security that they feel and be able to process a lot of those claims very, very quickly and be able to recover those funds and be able to for customers. I think that's one of the biggest thing that we have done in the past year and a half. Disputes absolutely important. I'll take the front end of it and it's around originations. And it's how can you look at all of the credit risk decisions, the fraud decisions and the compliance decisions at a holistic level use generative AI in the form of co-pilots to help guide resources through the origination process. Awesome, let's do that. And real quick, give a quick plug for your team and your group. What are you guys working on? What are your plans? What are your highlights for your group? Yeah, one of the things, I mean, team's absolutely amazing. I mean, we are driving innovation like you wouldn't believe. It was announced yesterday, our CTO's favorite solution is SaaS fraud decisioning. That is the future. It's laying down that architecture to make those enterprise decisions. It's got real, you can quantify the value proposition there. 100%. Money's involved. Thanks guys, appreciate it. It's too long. We'll be right back with more coverage. I'm John Furrier with Dave Vellante. We'll come up with the analyst angle up next where the industry analyst roundtable will break down what's happening in it and generate AI, SaaS, their customers and the industry landscape. We'll be right back after this short break.