<ul class="font_8"> <li> <div class="font_8">Blockchain implications for auto finance</div></li> <li> <div class="font_8">Unique ways to leverage location data</div></li> <li> <div class="font_8">Novel payment tools</div></li> </ul> [toggle title="TRANSCRIPT"] <div class="transcript-scroll-box">JJ Hornblass 00:00Okay. Good afternoon, everyone. Good afternoon and welcome. So great to see all of you here in San Francisco at this auto finance innovation. My name is JJ horn glass and I'm CEO of royal media. super happy to see all of you here. And looking forward to a great conference. We've been working very hard on it and really excited for it. It's the topic of the day, I guess, and just hope you have a wonderful, wonderful experience here. We start our events, with charity as so many of you know, and we are pleased starting this year to support the Absa Education Foundation. The mission of the American financial services associations, Education Foundation is to educate consumers on personal finance and to help them realize the benefits of responsible money management and to understand the credit process afsa. EF is an affiliate of the association and we're proud to make a donation to the foundation on behalf on behalf of the speaker faculty. So I want to thank the speakers for allowing us to make this donation and I want to urge all of you to visit fsaf.org to learn more about this great charity. I also wanted to thank our sponsors, we are greatly indebted to them for their support and helping us bring this event to you. And that starts with our diamond sponsor, auto gravity. And as well our Platinum sponsors, Deloitte and market scan, and also to alpha comply. Five serve Island Dix people strategies Inc. and white Clark group. Let's give them a round of applause for their support. 02:02 So JJ Hornblass 02:04 this is a technology conference. So I'm gonna highlight the, the hashtag, but I'm gonna expect you to use it this year, not like last year. So the hashtag is fy 2018. will be tweeting to that hashtag and following up with posts on auto finance news dotnet mainly after the event, but look forward to your use of that hashtag. We'll send you the presentations that we're permitted to share after the conference, so don't worry. And also, if you're in the workshop, there was a workshop this morning. That presentation will be shared as well. The Wi Fi network that we're on here is titled Park 55 meeting. And the password is a fy 2018 03:00 Okay, everyone got that. JJ Hornblass 03:04 Our next event is the auto finance performance and compliance summit, which takes place, March nine to 10. in Dallas, that hotbed of auto finance. And you got a postcard that allows you to share a discount with your colleagues, you're also invited to go to the registration desk here and get a lovely 15% discount as well. And it'll be very convenient for you to register. So I encourage you to do that. The mobile app, a lot of what we have a lot of information on the mobile app, all the questions we take through the mobile app. And there are two. There's there's a demo here. Let me explain how to get to the mobile app first. There are two ways you can go to the App Store and search events mobie event m o bi and our code is a fy 2018. And that would be for the native app. And if you'd like to use the web app, you can just go to event mobi.com, forward slash, fy 2018. So here's a little demo starting on the technology demos already. In in each. If you go to the agenda, and you find the particular session that we're on, and you choose that session, and towards the bottom of this session, there'll be an ask a question, option. And please do that week. We really encourage you to ask as many questions or make comments as you'd like, this is important. And and hopefully it'll make your experience here. As exceptional as it can be. Also within the mobile app, there is Something called the check in challenge. So at various sessions and also with our esteemed exhibitors, you'll be able to get a code to check in. The more points, each time you enter a code, you'll get points, the person who tallies the most points, wins series three Apple Watch, and which is not a bad thing. So I encourage you to use it. It's one of the options on the mobile app and do it. It's fun. I think 05:33 it's fun. It's fun. JJ Hornblass 05:35 Okay. So last week, I don't know if anyone noticed this, but last week, there an autonomous vehicle startup named Aurora raised $90 million. So if you're not familiar with this particular deal, don't feel guilty. So while $90 million is not exactly chump change, by Measure. There is so much going on in auto innovation today that it's easy to gloss over a $90 million deal. When I look at any autonomous related funding these days, 90 million or otherwise, my thoughts immediately turn to the auto financing. Each deal seems to angle the industry toward more auto finance demand or at least different demand. Maybe not today. But eventually, the innovation side of the auto industry that is connected cars, autonomous ride sharing and frankly business models, yet to be discovered, has captured the imagination of not just the traditional auto industry, but really consumers the world over. This is a wave of innovation and technology development. That will overwhelm the auto finance industry of that I am sure, just look around you. This is just the third time we've held this event and this year's auto finance, innovation, attendance has exceeded all our expectations. The obvious reason beyond the hard toiling work of my colleagues to present you with an exceptional conference, is that all this auto innovation is coming to auto finance. Certainly, it's difficult. It's difficult to conjure, what auto finance will be like in five years or even 10 years. But actually, that is our challenge here. That's the challenge during auto finance innovation, we're going to see demos from startups. We'll hear from exceptional thinkers from both the traditional and the startup sides of the industry. And Really there's just an exceptional group of people here diverse from all walks of our sector here to network with and learn from and I want to encourage you to do that. We're going to start with our first session, and I'm pleased to introduce my colleague, Natalie madula, who will spearhead it Natalie is the deputy editor of auto finance news, and its sister publication power sports finance. She joined royal media in 2016 and has quickly climbed through our ranks. Upon her recent move to add until her recent move to the Big Apple, she was an editor at the daily Sentinel in East Texas, and she's a wonderful member of our team and I hope you'll join me in giving her a rousing welcome Natalie 09:00 Thank you, JJ, for that introduction. And good afternoon, everybody, and welcome. Really excited to be here with you all today. We have a really great, great bunch of sessions on tap for you today. And it's all going to start here with this presentation entitled 10 innovation ideas for 2019. So innovation was once considered a buzzword and auto finance, which for many it still is, but I feel like it's sort of become more of this staple topic of discussion for for everybody, both large and small corporations. But are lenders embracing true innovation? I think that's a question that still stands to reason. The opportunity to innovate inherently revolves around key disruptors, autonomous vehicles, new ownership models, the blockchain artificial intelligence. These are just some of the many disruptors in the auto space that lenders should keep on their watch list. And joining us today to give additional insight into the innovation ideas that should be on your agenda over the years to come is Andy Clemens, Senior Vice Premier President of Information Technology at global lending services. Andy leads application development a global lending services and analytic and technology driven subprime auto finance company. Andy is a progressive technology executive with more than 20 years experience in it prior to joining global lending services, and he was at Capital One for 13 years where he built and ran substantial portions of the data environment including Capital One auto finance data warehouse, and he holds a Bachelor of Science in chemical engineering from the University of Virginia. So everybody, please join me in welcoming the end to stay sick and take it over. 10:43 Thanks, Natalie. 10:47 So I've got the challenge of talking to you for 30 minutes about 10 innovation ideas, that is not very long at all. And so we're not going to get in deep to all of these of course, but we are going to talk about some things That are right on us right now that we can use every day we can walk out of here and implement these things. And we're going to talk about some of the things that are are still five or 10 or 15 years down the road. But before we start with that, I'd actually like to ask you to indulge me for a minute and we're going to talk a little bit about art. So 10 years ago, I had a chance in Dallas was based in Dallas for a while to visit the Dallas Museum of Art when the jam w Turner exhibit rolled through town hall. Turner is an English artist who painted about 200 years ago during the Romantic period. And you know, as I went through the museum and saw these paintings, there really was some just majestic pieces of art. And what I want to do is show one of them here. Next I want to ask you to do is for a minute, suspend your analytical thinking, suspend the intellect and feel this painting, really sit back and say what is this painting? You know, how does this make me feel as a human 12:06 Things you want to call your attention to is the the blackness the darkness of the painting 12:13 the size of the sky, right? It's it's, it's more than half of the painting is sky. And in person, this is a physically impressive painting it's well it's over three feet big. I mean, it's it's a large, it's a large painting, you know, the roiling sea that the fishermen are on as they're trying to go about their work and catching fish and earning living for their, for their selves for their communities. You know, there is land in this picture, but it's this thin little strip of rock here. It's actually rocks off the Isle of Wight. And certainly they do not look inviting at all. And then there's a little bit of light on the scene. I mean, just just a teeny little bit just to illuminate the fishermen. But when you look at the clouds, it's hard to tell but for me, it doesn't necessarily look like the clouds are parting to let more light in. It looks like they're ominously close. In to leave the fisherman in total darkness. So why am I talking about this? Well, I think like the fisherman, we are on a sea that's ever changing. It's roiling. It's moving around us. You know, for me when I saw this, the emotion that it inspired was terror of man's insignificance to nature. And it wasn't till actually I went home and I did some research that I found out that this is exactly the emotion that Turner was going for 200 years ago when he painted this picture, which really, for the first time opened up more about the power of art with this, you know, the scan 200 years ago was in trying to instill this this image, what he called the sublime, which is man's insignificance to nature, and actually posit that for us, we're very similar to the fisherman. The technology that's out there is way bigger than any in any of us can understand. It's way deeper than we can get to it in 30 minutes, and in two days, or in two weeks, or even two lifetimes, we can get to all of it. However, what I want to do in the next 30 minutes is really try to park the clouds a little bit and We're gonna spend some time so that as we go through the next two days together in the hallways at lunch, over drinks at dinner, that we can understand these topics a little bit better. And like the fishermen, we can figure out how do we, you know, take this value back and and struggle with these but earn a living, you know, doing this and add value to our customers add value to our companies, and for ourselves. Alright, so the first one I talked about is Alternative Payment channels. This is one that's here, it's right here today with us. And when you say alternative, there has to be something that wasn't alternative. And I think originally that must have been, you know, cash and cheques. But clearly now to call online bill pay an alternative payment channel. I think that's that's laughable, right? I mean, I'm sure that most of us in this room, in fact, even as a subprime organization that has a lot of consumers that are underbanked, our predominant method of acquiring payments is through online bill pay. And of course, there's other ways you can get payments made out there. You've got texting, you've got in store so folks can go up maybe to a service center at work. Walmart, a service desk at Walmart, they can go in and present their bill make a payment, there was certain providers, you know, there's Facebook Messenger. But even more exciting than that there's some some innovations recently that are opening this up even further. One of those ones that I think is really exciting is the mobile wallet. Right? So being able to present the bill to the consumer right there on the phone, and what most most of us do, we wake up in the morning, the first thing is we roll over, we grab that phone, and we look at it. So if you can look at that and say, Wow, global lending services my bill is due today. I mean, that's really powerful, especially when you have maybe a consumer who's who's less skilled in managing their money being able to present right there on their phone that their bill is due in a way that that is absolutely defendable from from a consent standpoint, and they've chosen to put the app on there they can choose to remove the app at any time. They can see that bill, it's it's very friendly. And then of course, the other one is in lane. So in addition to being able to make payments at the Service Desk, why not be able to take your phone as you're checking out if you're especially for an under Bank consumer and you don't have a checking account, you've got cash and you want to be able to pay your bill, how can you go and pay that bill in lane just like maybe you're purchasing a credit or a cash card. And there are a number of partners, probably some of this room that actually have this feature and can provide it. So what's the challenges here? Well, the challenge is actually fairly minimal for this, really, the challenges are about integrating these platforms with your accounting system, and with your servicing platforms. 16:27 Now, the next one, this one's huge. I mean, huge, huge. In fact, I think this is the one that future generations will convict us of, you know, if you think back and you look at what the world was like before penicillin, and you think about the number of deaths and you know, childhood childhood mortality rates. Well, this slide actually talks about United States stats, but across the globe, 1.2 5 million people a year die in car related incidents. I mean, that is that is a staggering number of folks, I do think my grandkids are going to say I can't believe you ever got on the road, you got an A vehicle. Why did you get in a vehicle? Like that's, that's ludicrous. Why would you do that? It's like, well, that's just that's just life. That's we were used to it. Right? The United States is actually fairly safe compared to most other parts of the world. And 90% of those accidents are caused by humans, at least fully or in part. So what are we doing about this? Right? I mean, that this feels like a call to action for us in this room and for our partners outside of this organism outside of this room as well. This is going to be incredibly beneficial for humans, it's also going to be incredibly disruptive. So the slide indicates, you know, one in nine jobs in the United States is directly tied into autonomous vehicles. Right, whether you're talking about the Lyft driver, whether you're talking about the tractor trailer driver, you know, maybe a fire truck driver, right? Certainly, factory workers autonomous vehicles isn't necessarily just on the roads. What about the robots at Amazon that moves the freight within the factory? What about forklift operators inside of outside of factories or warehouses? Right. So we think about autonomous vehicles, it's going to be incredibly impactful. But it's going to add $7 trillion to the US economy or to the world economy by 2050. That is a mind boggling amount of money. So there's tremendous opportunity out there, but it's also gonna be incredibly disruptive. And as societies we go through this disruption, it's gonna be a lot like the Industrial Revolution and what we're gonna have to do to find and get jobs for folks that their jobs are being replaced by this. However, the disappointing thing, at least disappointing for me is that by 2030, McKinsey's only estimate that 15% of new cars will be fully autonomous. And that's, I think, a little a little disheartening, I'd love to see that come along much faster. And I'm sure that you know that, as you mentioned earlier, that you know, there's folks out there right now working on putting the bringing these to market, we're gonna see some cars out there. My wife has a Q seven, it's a great car, and it does a lot of lane assist. And of course, Tesla is great with this, but you know, it's just going to get more and more advanced. Alright. blockchain so we talked about this This was one of the topics in the session this morning some great conversation about blockchain. You know, blockchain is really a distributed ledger. And it confounds me to this day how I'm here at this conference. And if I realized that my dog is running low on food, I can pop on my phone with a couple of clicks. So I got a dash button that's for the food that my dog eats is a chocolate lab, he likes to eat a lot. And in two days, that food will be shipped from a warehouse into a truck and then delivered to my front door. However, if I go on my phone, and I try to move money from my e trade account, to my bank account, it won't be there in two days. It takes longer than that, which seems ludicrous given that we've got technology to move physical products that we can't move the digital products and why is that? Well, it's because we actually have to move as that money through a clearinghouse, right? There's got to be a trusted provider in the middle that ensures that there's no fraud going on. blockchain is a way for us to remove that the strip that clearing house in the middle, right and that's going to speed up that Time to market. So these transactions would be much closer to instantaneous. That is what we'd expect in, in today's modern environment. And this is one that having, you know, worked to Capital One, and I'm sure a lot of folks in this room that worked for some of the banks, you know, it is it is frustrating, like, how come we can't make these things move quicker now we understand why that is their security, and there's fraud, and there's things we have to protect our customers against. But we can't use that as an excuse, right? We have to say, how do we get past that? How do we use blockchain to really blow through those barriers to deliver a fundamentally different environment than than we do today? Of course, in auto finance, the use cases are around you know, I think vehicle ownership, how do we track the titles, there's a lot of work with paperwork that goes around both internal company as well as with the state governments, with our consumers. You know, we got to get signatures on things. It's a messy process. It's different by state, you know, being able to use blockchain as a way to to streamline that I think is gonna be very helpful. Now, there are some big challenges here and this is one we were talking about earlier. I think in many ways, you know, blockchain reminds me of the Shift Tab definition Television. You know when that when the HDTV shift came along, you know, you had to get consumers to shift, you had to get the government to shift, you had to get private companies to shift. And I think blockchain is very similar, you have to have a number of entities moving in lockstep to be able to make these changes. So this one is coming. I think it's probably closer than autonomous vehicles. But I don't know that this is next year. I mean, I definitely there are already folks out there using this for securitizations and some other mechanisms that they're trying on the kind of the bleeding edge. But I think broadly for the folks in this room, this is a few years down down the road. 21:30 All right, this one's fun. Machine learning and AI. I mean, those are different terms. But for purposes of this, I'm going to refer to them as one thing. And I love this quote that's up here. And when we think about machine learning and AI in the future, you know, there's often choices and we're thinking about it, do we have something like this? Which you know, one of the best movies of all time, I have to include a matrix reference. You know, even though I love the movie and Neo is a great hero. I would not Want to live in that world? Right? That is clearly a technology dystopia. You know where the other angle is, we live in something that looks like this, which clearly is a technology utopia until actually right before I came in here the other day, we just submitted the slide decks a couple years ago, a couple of weeks ago, I realized that the person walking the dog actually has like his arm extended like a robot arm. So I'm not sure that there's actually a human in that picture. So 22:26 that says something about our human our utopia the future. 22:32 But for us in 2019, what does machine learning and artificial intelligence Well, I think there's a couple of main reasons and main use cases for artificial intelligence right on us right now. One of them is classification of the present. So how do we use artificial intelligence to take some of the work that our teams do that are really you know, the the routine the rote work, the things that you know, aren't aren't fun or sexy? Is the contract signed to these names match, you know, being able to use artificial intelligence to work remove some of that tedious work from our workforce and improve our efficiencies also improve compliance and scalability. The other one I think is much more fun is prediction of the future, right? A lot of us obviously, it's we're making loans, we're doing it because we have some belief about what the future loss expectations are going to be. So being able to predict the future more accurately is something that's extremely interesting to all the lenders in the room, right, the better we get this done, the the, the more we are going to profit by. And so I think being able to use artificial intelligence for the purposes of, of pricing loans, as well as for behavioral models and collections, and how we how we can ensure better collections. Now, the challenges with machine learning, and actually, I'm going to go through these more one at a time, because I think this is very important. The challenges of machine learning are not that different, in fact, very similar to the challenges of traditional modes of predicting the future, right. And so ultimately, at the end of the day, when you're thinking about using data to predict the future, you have to ask yourselves well, to what degree and what quality is my data and how Complete is my data do I have? Do I have really all the data points that I need to be able to make an accurate prediction? Or am I missing some key elements of my data? Has my data been censored in some ways because of maybe previous treatments that I've done that prevented me from collecting some of the data that I need. And sometimes you if you're not really careful, you can find yourself making terrible decisions, because you then in retrospect, to realize that, that your your algorithms were actually censored by previous treatments you had already done. This next one is really tough, you know, think about chaos. And what I mean by that, if you think about political, social technology events that happened 911 you know, might be one, the 2008 market crash, there are things that happen, where our models in the data you have of yesterday does not reflect tomorrow, right? There is a shift that happens and all of a sudden the world is different. And so just like in traditional model builds machine learning has the same problem. How do you predict the future when the future doesn't look like the past does and we all have to live with that uncertainty. And then of course, the bigger one for machine learning. I think there is some of an answer for this is the technology, right? So there's still it takes a lot of expertise to be able to do the things that even forget machine learning, just to be able to solve the challenges we just mentioned in a regular way with with statisticians and with with data scientists, but throw on top of that the limited knowledge of the skill set that's out there for data scientists and bringing in folks that have experience in building machine learning. I mean, it's it's tricky. And the funny thing is machine learning isn't new. I mean, 20 years ago, I was playing around with some algorithms that were called genetic programming. You can add neural nets 20 years years ago, in the book that I was teaching myself how to do this on was 20 years old at that time, right. But we're now just getting to the state where computers are fast enough. We have infrastructures of service, which we're going to talk about later, we're getting to the place where it's now feasible to do machine learning. Alright. 25:52 I was like, I hope that's not mine. 25:55 alternative data. So, you know, when I talk about alternative data, really what I'm proposing In this room, what we're going to talk about is anything other than traditional credit bureau data. And a great example is, you know, you get utility bills, location data, the cell phone bills and utilization. How do we use this? Well, we just talked about machine learning, we talked about one of the challenges with machine learning was how do I have his broad database as possible? How do I have great data quality, and being able to leverage alternative data is one of the ways that whether using traditional model builds or machine learning that you can actually, you know, enhance the quality of the models enhance the quality of the of the prediction. Now, some of the challenges here, I mean, certainly FCRA, if you're going to use any data source to make a credit decision, you wouldn't want to have a consumer impacted because you have bad data. So you've got to have a high quality source of data if you're going to use it to make In fact, if you're gonna use it for any prediction capabilities, you want a high quality source of data, but especially true when you're making approve decline decisions. On top of that, if you're going to make an approve or decline decision, the consumer has to have the right to dispute something on that that they don't think is is accurate, right so they have to Be able to dispute it and looking for data sources, you know, that have these these, these hallmarks is very important as you think about alternative data. The other one I mentioned is hit rate changes. Dan Bankston, our chief risk officer, you know, brought this out to me said, Listen, you know, we've got these great partners out there, they're building this alternative data. But as they go about enhancing their product and bringing data in, it actually fundamentally changes analysis that you may have done six months ago about, you know, how you're going to use it your model and as the data sources constantly evolving and changing which it needs to to continue to grow and add value, then what just happened to your model as if the data that's being supplied to that alternative data source provider has changed and it's something just got to be aware of and update the model refreshing 27:46 location services, so I have a really guilty secret here for a minute. This is terrible. I know especially coming to San Francisco, but prior to this week, I had never used Lyft or Uber 27:59 but they are Fantastic services. Their secret? Oh, 28:04 yeah, I'm in Greenville, South Carolina. So you know, we drive our cars everywhere, right. 28:09 And so you know, I put this up here because I do think there's some fantastic use cases for location services. I mean, Lyft and Uber are one of those. Whether I'm I love the fact that when I opened my weather app, and today, it tells me what the weather in San Francisco is it knows I'm here. I'm not in Greenville, South Carolina. I think there are some use cases for location services in our industry. I think particularly one that I mentioned before is if you're going to have an underbanked consumer walk in and make a payment in a Walmart or CVS, being able to have that consumer and know where they are and tell them here's the here's the nearby centers where they can take that cash payment in and make that payment to you. I think that's a fantastic use case. If you're maybe a large organization, you have more than more than auto lending as your product, maybe there's some cross sell opportunities that you have there. Maybe the customer has their car breakdown. They want to know when nearby service centers are sure those are great. However, This is the one that I think I'd put the most caution around and that's why I put this little sign up here you know we see you keep that in mind. I think certainly that's us to the consumer but it's also the consumer to us right? Hey wait you're looking at my location data Why do you need to know where I am? What what valid use case I mean, I got it why why Lyft needs to know where it where I am but why does my auto lender need to know where I am like that seems kind of creepy. And so I'd really just asked you know, offer a word of caution here about really making sure that when you're using this as something that is going to be consumer friendly you know, these things are creepy anyway I mean, I you know, my boss loves to visit tell your ad and and he said he got off the plane and Facebook said Hey, welcome to tell your rod. It's like ooh, like yeah, that's that's kind of creepy. You know what why does Facebook need to know right now where I am is that really that are really want everyone to know that and so just just as we use location services, especially auto lending, I just say Be very careful with with how we think about this great technology but certainly open to abuse. Alright, so big data. Which I kind of hate this term, but it is the term that we're going to use is going to help us solve some challenges that alternative data actually creates. Right? We use defines our loan originations provider, and we've chosen to get a big extract from our LMS system. Every time a new application comes in the door. Every time one of our funders or buyer saves an application, we get this large chunk of data. There's thousands and thousands of attributes in this data file. So how do I take that in and make it available for analytics? Well, one of the things I could do is I could take it, I could parse every attribute out. But that seems like a huge waste of time. You know, we want to have a data warehouse that's organized in ways that makes it easy for analysts to see all the data that they need on a regular basis. So let's take that XML and let's parse out the elements that they most you know are going to use, and let's put those in easily consumable structures. And then what I want to do is be able to take the rest of that and store it somewhere and big data gives us the ability of doing that, right. I want to store that in a data lake. I'm going to put those XML files in the data lake. And if an analyst comes along and they say hey, ain't I'd really like to have this tag that I now know is super predictive. But I didn't know it when we first started, you know, pulling the data. And I didn't know when we first started passing it out. We can take Hadoop or hive or you know, Hadoop and hive, maybe you might use spark or other in memory, you know, optimizations on the Hadoop platform, but take some of these technologies and grab those XML files and parse them. before when we used to have this problem. Just two years ago, we were much smaller. And I recognize compared to some of the folks in the room GLS is still a small organization. But two years ago, when we were much smaller, you know, it would take us months if an analyst came along and wanted an extra, you know, piece of data out of the XML files, and we had a tiny fraction that we have today. And if I had to go in today and write code to get those data out without a big data platform, I'd be spending maybe a couple of days or a week writing the code. And I'd spend literally months trying to execute that code against traditional hardware to pull that pull that information off with big data, the ability to store that data and process that in in much quicker turnaround times. I mean, hours instead of months, literally is something That we now have at our fingertips. 32:03 One of the things that makes Big Data even much more effective, though, is Infrastructure as a Service, this one was just so cool for me when we first started using AI, as, you know, the ability of going on and using infrastructure, like a light switch. It's like, hey, an analyst comes along, and they want this extra data element, I've got a hive query that we write takes a couple hours to write, and we go ahead and try to run it. And my developer comes back and says, Andy, you know, it's gonna take us still a month, the parts to state of the day. And once it's like, what you mean, it's gonna take us a month, it's like, well, you know, we've got our regular Hadoop cluster that we run our nightly stuff on. And so we have to run this, you know, off hours when we're not clustering. The clusters this big is built for a certain process. It's like, Well, why don't we just stand up another cluster? And it's like, oh, yeah, I guess we can do that. So if you're, if you're using Google, if you're using Amazon Web Services, if you're using Azure, you know, standing up a Hadoop cluster is I mean, seconds, right? And you can do it through a PowerShell script. You can do it on a schedule, you can scale the clusters up and down. That's something if you have your own data center, you're And that stuff internally, it takes, I mean, I know it Capital One, as we were building out our first Hadoop cluster, I think it was like a year long project to get that thing in place. Maybe not the first prototype, but the first one we're really using. And so I mean, I think from from Infrastructure as a Service, you know, this is just phenomenal. In fact, this graph kind of illustrates, 33:19 really the power here. 33:22 You know, when we do this, and this, you know, the later part of this curve here really shows that as we scaled our nodes up as we scaled in the each node in this case had four cores on it. And so you can see that our last test case, we ran 192 cores, you know, literally we were sitting around on a Wednesday afternoon saying, okay, we need to parse this data out for for our analytics team. And we just started running a day's worth of data, getting an idea of how many miles we could process in a minute, and then started scaling the cluster up, you know, trying out different, different settings. By the time we got to 192 cores. It was about seven o'clock at night. We said Okay, you know what? Like, it's going to take 12 hours to let this run, let's just go ahead and let it run, we could scale it up faster. But let's just go home right now it's fine, we'll let it run and we'll come back tomorrow and all our data will be parsed out. So you know that the ability of turning on infrastructure at the flip of a switch, I think, is super powerful. 34:18 All right, so Software as a Service, then, 34:21 you know, one of the models that we've taken in turn our organization is to really leverage Software as a Service, we have a very small tight development organization. And the way that we do that is is really kind of thinking about buy versus build and leaning into by whenever we possibly can. Right. And of course, you know, that this this, this is very popular these days. In fact, there's plenty of folks in the room that have software as a service to sell, and being able to take these products and integrate them into your, you know, custom framework is extremely powerful. And whether that leveraging machine learning, right, or whether they're leveraging alternative data, big data. You know, most of these folks use is on the back end, but being able to to leverage these off the shelf Software as a Service, I think is extremely powerful. Or if you've got something and we do like our proprietary structuring and pricing algorithms, we actually built as a software as a service, you know, piece of code, so that we can actually plug it into our LMS platform. And so we effectively have built our own internal Software as a Service, even though we bought an off the shelf lls, and we basically call our software as a service for the proprietary things that we want to code up and develop inside. Now, there are some challenges with using software as a service. I mean, certainly internal resistance, you know, I mean, for especially for larger companies that have been around a long time, you know, that that buy versus build mentality can be shifted and leaned heavily into the bill. It's like, no, we're unique. We've really got something that's special here. And no one else doesn't like our do, you know, only us can send statements the way we send statements. No one does credit bureau reporting like we do credit bureau reporting. You know, whatever it is right, we can get a little tied up and wrapped around the axle on some of these things. Now, there are things That certainly makes us all unique. And that's why I'd say build that stuff in house, but leverage software as a service to really give yourself you know, power and speed and flexibility. There are some other challenges. Not that anyone that room you know, represents this, but there are some things out there that is vaporware, right. I mean, there's some there's some providers out there that really you know, they've got some great ideas, but it's it's not really in in place yet. So really being able to, to find a great partner to do a solid integration to get over some of those internal resistances to have great vendor management, you know, these are all kind of keys to leveraging software as a service. 36:39 integration. So this is number 10. This is the final one. And I'm going to use you know, this device I think is the poster child for integration. So you know, over the Christmas holidays, I how many you guys have an echo.or an echo or echo show wearing this thing's for quite a few folks. So over the holiday You know the echo dots about 50 bucks but Amazon especially say you can get one for $30. So we had a white elephant gift exchange with a family we actually bought two echo dots as a presence. My wife and I didn't have one we said, You know what, they're 30 bucks one I will buy one and put it in our home. So we got it under playing around like man, this thing is really neat. Like you can have Alexa tell you the weather, right you can have you can have Alexa, go off in and order something from Amazon, right? You've got your I've already got my Amazon Amazon's a software as a service that you can go out and, you know, Alexa, the Echo Dot reaches out talks to talks, Amazon is able to purchase products for you using your voice. And you think about the dot like what is the Echo Dot like in terms of hardware? It's it's a microphone, it's an LED? It's a couple of buttons. It's an internet connection. It's basically nothing right? I mean, it's it's so small, but the power of it is all of the integrations that it has with all of these other software as a service as many of which already existed. Right? They're not brand new. They weren't created for the use with the echo.or the Echo, but they're being leveraged About this to really pull all these things in one place. So as a consumer sitting at your house, if you want to buy something, you can buy it with your Echo dots in there. If you want to know the news, Alexa can tell you what the News is. You know, if you want to turn your lights on and off, see who's at the front door, I've actually recently got into doing some home automation and have learned to do a little bit of wiring in the house. I haven't bought anything down yet, but you know, installed some of the some of these great devices, right and be able to turn lights off in different rooms sit down to watch a movie, tell Alexa to turn the lights off, just you know really makes everything handy, right? It's not the tip of your fingers, like the phone, you can be having your hands full, and still be interacting with this technology. Of course, audible being able to play a book and then if you want to play a video game, maybe and interact with some of your friends, you know, Alexa is able to help you do some of that as well. And when I put this up here, it's because you know the power of this device. Yes, there is a great software problem. The back end it's a text to speech and the speech detect peace and understanding the human I mean that that that's tremendous. I don't take any other thing from Amazon for that. But more than that, it's their ability to integrate with existing other software providers that are out there and to provide all those services, you know, through their device. And when I think about us and lending, that's one of the ways that we try to build our systems, right, I want to have a central platform within my servicing platform originations platform, I want to be able to tie into alternative data, I want to be able to tie into machine learning, I want to be able to tie in, you know, to all these other products that are out there so that I can leverage them to make our institution better and not be limited by how much stuff that we can push through in a year or our internal resources whatsoever. These, you know, these these great partners that have built these things that, you know, we're in some ways we're sharing knowledge, right? So the difference between Linux versus Unix. It's the ability of taking the best and brightest and having them work together to create solutions, as opposed to always competing on things that really aren't differentiated competitive advantages for us anyway. So let's work together on them. And that is it. Any questions? 40:07 Alright, sorry, I still can't get over the fact that you've never taken an Uber before today. So 40:11 I still never taken an Uber. I've taken a Lyft 40:13 Oh, I see. So I'm from New York, that's all we do is take rideshare. So, so you talked a lot about machine learning in your presentation that I thought was really interesting. I think I'd probably be helpful if you could share with us like, sort of To what degree Global Learning Services is currently using it? Or if you're not like, what ways you're currently like really exploring right now to 40:32 use? Yeah, yeah. So we actually use it a little bit in our in our model build process. So we don't actually have it in production with the model running and daily approved declines. But if you think about trying to parse through all the data that's available and figure out what which of these, which of these elements is really predictive. Which one better slopes risk? You know, certainly we could have traditional statistics techniques that would help us find those and we do. But on top of that, being able to leverage machine learning to help speed up that process is one of the things that I think is very beneficial. There's a number of things that we're not doing. But we're actively engaged in trying to figure out how to make happen. And one of those that we got a couple of things going on, right this year, one of them is in our funding department really looking at some of the classification stuff that we said. I mean, if you think about our businesses, we've been growing so dramatic over the past few years, you know, how do we hire enough people in and then train them quick enough to handle right now in the middle attack season, right? In subprime tax seasons, this huge wave of volume comes in? How do we train enough people up to be able to do you know, income validations, or checking to make sure the names match across the title form of the credit application or, you know, the, the contract and so being able to leverage machine learning in that space would be fantastic. And we're leaning into that some. The other one is in our collection space, you know, as we think about well, how do we do collection strategies and how do we, you know, build models there? Is there a way for us to leverage machine learning to help us you know, in that domain, 41:54 and something you mentioned to you said, it's hard to predict the future when it's constantly changing, so you can't wait Really, fully predict that. So you know what team roles are needed to help implement, you know, machine learning. So does it require you to have like this team of data scientists to really hone in and work on this constantly? Or what sort of the process better? Is it better to just outsource this? 42:16 I mean, well, so while there certainly are outsourced providers and consulting groups that will do some of that work for you. And then that's absolutely a viable option. If you don't have the expertise in house, there's number of providers that will say, they'll basically do that function, but they're gonna do that function with those roles, right? They're going to use a data scientist. In fact, that is one of the definitely pitfalls of machine learning. It's like looking, you can just throw data at it. But if you don't understand what you're throwing at it, and you really haven't thought through it, then the output that you're going to get, you're not going to understand either. And so being able to and this is the debate that that Dan, who's our chief risk officer and I have constantly is like, Well, how do we push the boundaries on machine learning, but still make sure we're, we're holding true to, you know, all of the expertise and all of the learnings that have been developed over kind of lifetimes of you know, data analysis and stats and I think there is We're starting to find ways to leverage more and more machine learning. But you have to be very careful about how you do it because of the pitfalls with, you know, understanding your data. 43:09 And you talked a little bit to you about some of the compliance hurdles when utilizing this technique. So, you know, while it appears that like machine learning could essentially be beneficial from a compliance perspective, because it kind of, like eliminates or sort of eliminates that human error. I feel like there could also be some sort of regulatory concerns there as well, because regulators often like to know how you got from decision a to decision v to decisions so on based on the data points, you know, turn in for that. So, you know, from a credit decisioning like aspects, you know, what sort of concerns Do you see there, and how do you kind of overcome those to show regulators how you got to that decision using 43:48 this machine learning process? Well, and that's where for us, we actually would use machine learning to help us pick the variables not necessarily in the production model, and so for any individual applicant to come through, we can show exactly how we got to that decision. Because we've got a more traditional, you know, scorecard and model built on it. And then how we come up with that scorecard is really that's where we can we're leveraging more machine learning now, you know, is there a place somewhere down the road where you would do credit decisioning using yet maybe, but I think we're still, you know, pretty far away from that in terms of approved declined decisions, 44:22 such as Symantec. A couple questions here from the audience. Many of these technologies are more generic meaning they're applicable to many sectors, not just auto finance, for example, blockchain location services, etc. are there technologies indigenious only to auto finance and what are the ramifications of them? 44:42 I don't know. 44:45 I mean, not not that I've thought of with this. 44:51 I don't think so. But I mean, maybe somebody's got one that that that we've missed, that is only auto finance related. I think everything that works Doing is really pulled off from, you know, shared technologies with other other industries. In fact, I think one of the things that has really kind of hampered us, if you think about a lot of other industries, I mean, you know, we'll have these discussions with folks about we're being innovative. It's like, well, we are for financial services. Right? I mean, we're not if you compare us to lift right, if we're not, if you compare us to Google, we're not when you compare us to Amazon, like, there's, you know, most of the companies in this room like aren't on the cutting edge, we're on the cutting edge for financial services, right. But to some degree, that that makes sense, because we have regulations, and we have FDIC, and we've got, you know, a ton of compliance concerns and we've got, you know, consumer data and so, you know, for us it does take a little bit more to push the boundaries means taking all of those things that we need to be careful of and not break, not break the bank, if you will, and then advancing the ball so we can get some of these products and features out to our consumers. 45:53 Your blockchain question blockchain question for you. So blockchain seems as though it will unwind many protocols. We have spent a lot of Time and treasure developed me, they work well, why would people be willing to mess with them? 46:05 I mean, I think ultimately it's it's it's speed. I mean, it's it's back to why do I want to wait two days to move money from my ETrade account to my bank? You know, it doesn't make any sense that I should have to wait that long. And customers don't I mean, you start using things like lift and you realize, wait, I can get I know right now and get a car in one minute. And I get you know, Amazon's I mean, you've got other things outside of auto finance, that are going to cause the consumers to say this doesn't make any sense. I want us to take so long for these things to happen. I think that's what's going to be external, external pressure. And I think internally, as we're pushing it, they're going to be some folks who say, wait a minute, we can get a competitive advantage amongst other folks even in this room if we have an implement some of these because we're going to make it easier for the consumer, we're going to make it easier for the dealers. And that's going to propel us forward in a very good way. Right? This competition will help us move move forward. 46:53 Cuz I'm more of a forward looking question that I think is probably on a lot of people's minds. So in autonomous will the OEM or the software Manufacturers own the consumer financing relationship and why? 47:04 That is a great question. Yeah. Right. And I think there's a great question about, you know, do consumers still buy cars? Right? The dealers just own cars? And does everyone lease a car? I think there's there will be does that matter? Why do most consumers still buy a car in a dealership and not online? I think some of this is rooted in and you know, possession. And as Americans, I think we like to possess our vehicles. And so I honestly if I could, you know, still live in an area, maybe not to live in San Francisco. Because the parking is ridiculous, expensive traffic is horrible, but sorry for folks that live in San Francisco. But you know, for for most areas of the country, they're a little bit less urban. You know, being able to own your car and keep your junk in it right and have your presets on your stereo and be able to have a jacket in the back. You know, I think some of those things, it's like, no, this is my car. It's, it's if it's dirt, it's my dirt. You know, it's, I think there's still that mentality that's going to be there. So, you know, it does it become more of a sentiment thing, versus a call Like, even if you could lease a car or rent a car or rideshare, car 10 or 15 years now, would you? You know, I don't know. And that's where it's, it's I think it's fascinating because I think that that autonomous vehicle, you know, could wipe out our industry or it could can make it, you know, much more important. It's, it's a, it's a tsunami hitting towards us and we're in that boat. But what form it's going to take, I don't know yet, and I think we just really need to keep our eyes on it. 48:28 So at the workshop this morning, there was an Alexa demo of loan origination, which to this person, it said, seems really complicated. So how can the complexity be overcome to facilitate voice originations? 48:43 Yeah, I mean, in his voice, originations, how consumers want to buy cars or not, I mean, I think that's it's a great demo. I think it's certainly a great proof of concept of what is technically capable with with Alexa. Once again, I think it's more about are consumers really going to go that route are they going to want to have you know, the car dropped off to them, so they can But at their home? 49:02 Yeah, I don't know the answer that man. That's a great question. 49:06 So referencing back to blockchain, so how can the auto finance space better leverage the blockchain? So it seems like, based on my conversations, anytime like I have a conversation with someone I bring a blockchain or just like, what is that again? Can you explain it? So it just seems like maybe there might be some sort of barrier there too, in terms of just understanding it is that maybe a big reason why it's been more popular in other spaces and not so much in auto finance or, you know, what are your thoughts around how we can better leverage that? 49:32 Yeah, I think the challenge for blockchain is really that we've got to get, you know, it's going to require government, it's gonna require dealers getting getting I mean, he contracting, I mean, contracting has been out there for a long time. And I think in a lot of areas, the contracting is really popular. But if you look at subprime, you know, even though the contract has been there, subprime dealers just do not, do not do it, right. And so what's the adoption for blockchain going to be and it's, you know, we've got to collectively move together. Gather a step at a time to get there. And we will. I mean, I think it's coming down. There's so much value that it's going to create that I think it's inevitable, but the pace of that play is going to be slower. I don't know that it's gonna be necessary because of technical understanding. And really, I think as a lender, I'm not trying to move if we use blockchain, I'm not going to technically understand how to use blockchain, right? We're gonna bring in a provider that's going to understand blockchain. And we're going to be tying it we're going to integrate it right. So I'm going to someone in this room or someone outside this room is going to say, Well, this is how you do it. And I'm gonna say, Great, here's how, let me tie that into my servicing platform. Right. Let me tie that into my originations platform. And then we'll end up leveraging it from that for most lenders in the room, I wouldn't expect that they're going to have to go off and really deeply understand blockchain unless they're determining No, we're going to build this in inside we're going to participate in the consortium. And I don't think most of us are going to have to to really leverage the the benefit of that down the road is if we're, if we're willing to buy versus versus build and integrate them in. 50:56 And so you talked a lot about sort of these AI innovation ideas for the world. Overall industry and everything like that. So what? For you particularly at Global Learning Services? What are starting your tech goals that you're really working towards this year? You already kind of mentioned a little bit of the machine learning aspects. Yeah. You know, other things that you mentioned this morning. What are sort of the goals or initiatives that you're undertaking right now? 51:15 Yeah, I mean, I mean, you know, I think certainly 51:19 if for us as a growing organization, you know, we're constantly thinking about well how do I how do I staff up you know, if I grow from double last year was double in the year before was a double and so how do you you know, and it's not going to be great it's gonna keep doing that but obviously it's gonna hit some someplace we can't keep doubling you know, how do you hire people in and train them and automation as an answer to that and so that's where machine learning comes into play. And then on the servicing side, you know, it's it is we don't right now have the the mobile wallet, right. I'd love to get that in play this year. We've got that on the on the radar for the year. We're able to accept payments through our provider at the Walmart you No kind of kiosk or service desk went on, say in lane, and we're trying to get trying to get that in and working with I provided this year to be able to accept those payments there. You know, we're really trying to improve our collections and servicing strategies, right, trying to try to figure out how do we maybe with a little lower touch achieve the same results? We've had some really good years in our collections and servicing space. But really, how do we kind of transform that to really use digital communications a lot easier, be lower touch the consumer? You know, obviously, I think it has some compliance benefits as well, you know, you're agitating the consumer less. And I think we can actually get the same level of collections with a lower touch strategy. And so we're really implementing a more advanced collection scorecard and making better use of that. I mean, I'd love to see us if you're, in fact, I was having some conversations this morning. Before that things started about how do we leverage machine learning to basically at the individual consumer level tailor a strategy for that consumer based on what they're telling us about their preferences and how they respond to our communications. I think there's a tremendous opportunity there 52:59 and the mobile wall Are you partnering with someone for that that we are? Yes. Do you have any timeline for when you hope to get that out this year, 53:05 we should be able to get in third quarter. 53:07 So we'll see. 53:08 Gotcha. Well, we're just about out of time for questions. I want to thank you and I want to thank Deloitte for sponsoring this session as well. So thank you again, Andy, for joining us. Great. Thanks. </div> [/toggle]
