Why Clean Data Matters More Than AI Tools In Collections | Ep. 2

Can AI really transform collections if the data behind it is flawed?

In this episode of Applying AI, Adam Parks sits down with Mike Walsh from EXL and Manny Plasencia from TransUnion to explore why data to fuel AI in collections is the real driver of performance.

They break down how data decay impacts AI decision making, why AI acts as an amplifier of existing data quality, and how organizations can strengthen data foundations before scaling automation. You’ll also hear practical perspectives on how to improve data quality for AI in collections.

Listen now and subscribe for more real-world AI insights shaping receivables management.

Podcast Website:
https://receivablesinfo.com/applying-ai/data-fuel-ai-collections-plasencia-walsh

EXL Website:
https://www.exlservice.com/

TransUnion Website:
https://www.transunion.com/

Mike Walsh LinkedIn:
https://www.linkedin.com/in/mike-walsh-b88b271/

Manny Plasencia LinkedIn:
https://www.linkedin.com/in/manny-plasencia-69562631/

Adam Parks LinkedIn:
https://www.linkedin.com/in/adamparks/

artificial intelligence in collections,
data driven collections strategy,
machine learning in receivables management,
data quality for ai decisioning,
right party contact optimization,
receivables analytics,
ai strategy for debt collection

#datafuelai #mikewalsh #mannyplasencia #aicollections

Transcript

SPEAKER_02 0:00

Hello, everybody. Welcome to Applying AI, the show where we cut through the hype and get real about using artificial intelligence in regulated industries. Season one of Applying AI is presented by Receivables Info and sponsored by EXL. So let's dive in. So this being our second episode of the Applying AI series, co-hosting here with Mr. Mike Walsh from EXL, because we have so many great conversations about artificial intelligence and actually being able to use it, not just theoretically. And so, as part of the challenges that I think the debt collection industry faces when it comes to artificial intelligence, is AI is an amplifier. And it's either going to amplify good results or bad results. And that's why I think that the data that we feed into these models and systems is going to be mission critical in regards to the success of deploying these tools. And so we asked Manny Placencia to join us from TransUnion because he's an operator and because he is looking at all of that data and trying to better understand how it can be used across the debt collection industry for agencies, debt buyers, law firms, and creditors. So, gentlemen, thank you both so much for joining today. I really appreciate you coming on, sharing your insights. I'm excited for today's discussion. Absolutely.

SPEAKER_00 1:19

Yeah, thanks. Thanks for having me, Adam. I know we've uh we've geeked out on various topics along the AI line, and I'm looking forward to diving in here.

SPEAKER_02 1:27

So I think one of the interesting things where I I thought I would kind of kick off the conversation is around the idea of AI being an amplifier. If we if we're di if calling out voice AI, if we're sending emails, if we're sending text messages, we can look at artificial intelligence from a few different perspectives, whether it be the content, the channel management, the um the coordination across multi-channels. And so there's a lot of different things that we can use it for. But as we amplify our ability to communicate at scale, I mean, if we're sending to if we're calling bad phone numbers, we're gonna call a lot of bad phone numbers. If we're sending emails to bad addresses, we're gonna do that at scale. And it's gonna happen quick. So as much as there's opportunity for us to grow and to learn and to uh increase our our tool usage, it also comes with a challenge and a threat. So, Mike, like what have you seen as you've been helping organizations deploy these tools across the space?

SPEAKER_01 2:26

So it's you're right. It's an engage, like the main AI that people talk about. Let's just put that out there to clarify, is right, is is engagement, right? Not not scoring or or or QA, but I mean the push right now is this is a better way to engage customers. It's faster, cheaper, easier, better customer experience. Let's just use this, this, these engagement tools to the math, right? I think why we have many here is not just running into bad data versus good data, but also enhancing data, using enhanced data. Where do we go with that? And then using data efficiently and right and and then figuring out data points that are predictive and become part of our new strategy. Like to me, this is the changing of collection strategy. Just like email and text changed collection strategy a few years ago. AI is gonna give you data that helps you use those tools or is using those tools way better than a human could. But as you said, without people like Manny giving us good data or their his clients giving us that data to use, we may be sending a lot of emails, virtual agent phone calls into outer space, into a cloud that goes no like nowhere. So, I mean, the beauty of AI is you'll find out if you have bad data pretty quick. It's not, it is not, it does not take long. And we do find some really funny data, like no.com email addresses, or even worse words. But you know, the a lot of the challenge we run into is as a as an engine or as a tool for our clients is it's their data. Me as an AI, you know, engagement tool, it's not my data. I'm like a doppelair for for engagement, but I it's up to agencies, creditors, debt buyers to make the decisions on how to use their data, what they're comfortable using. And I think that's a big discussion that a lot of people are not having.

SPEAKER_02 4:23

But it's also about the decaying value of data over time. So, Manny, talk to me a little bit about the value of data over a time.

SPEAKER_00 4:32

Yeah, absolutely. Look, we we're in it, we're first off, let me start out by saying we're an industry that's been primed for AI. Especially, you know, if you think about the the the the way our industry's grown and the way that we've utilized tools, not just to comply with the, you know, gosh, the growing landscape of uh of compliance and regulation that we dealt with as an industry, but also trying to grow from the what the industry was to the industry is, and you know, trying to meet consumers and expectations of the consumers, whether it be communication or payment, you know, whatever it is. That that's all evolved along with how we communicate in AI here to stay. I mean, I think we uh we discovered that what um um 73% of companies in 2024 were reviewing or exploring AI now, 93% are there. I mean it's my favorite stat ever, many. Yeah, I I understand. I understand it. And to your point, Mike, defining AI in your strategy is key. Um I think you know, I get excited about AI, and I've used this term before. I tell Adam all the time, yeah, look, this is the new frontier, everybody's excited about it. I'm excited about all the pioneers exploring this space because data is the new gold, and I'm a data guy. So it excites me, obviously, and the opportunities. I've been working closely with Adam, you, and everybody else in this industry to understand applications, use cases, and position transunions strongly to be able to support that. One of the key identifying factors that we utilize to explore performance is data decay. You know, the integrity of your data decays year over year, month over month. Behaviors change, people move, um, addresses change. I I get a new job, I make more money, I lose a job, I have less money. Family change. There are so many life events that change over the course of time that impact the data that maybe your creditor, you're relying on coming from a creditor, maybe you're even a secondary or tertiary agency. And at this point, you're trying to utilize that data to try to contact somebody or you know, segment somebody somewhere or another. And we've experienced that on a manual, in manual operations. Then as we move towards automation, and to your point, Mike, this is the new evolution. You know, it's not a trick, it's not a toy, it's a new evolution. And my fear is, and what I've seen is, you know, what was that stat that you and I talked about? It was 30-something odd percent.

SPEAKER_02 6:47

36% year-over-year data decays. And so if we think about how old the data is that we're dealing with from the creditor, the email address may be from the original origination. Then you have to consider delinquency time frame, charge off time frame. If it's coming to me as a primary, secondary, tertiary agency, I might be three years behind.

SPEAKER_00 7:06

And I have a unique, you know, I I want to say opportunity here at TransUnion because we're a credit bureau. It's not just that we're fortunate enough to have all our clients both on both sides of the house, right? Collections, as well as originations, creditors. Uh, we all work together in one big environment which is very unique. I know we're probably the only one of the bureaus that's has dedicated third-party collections, and we try to, you know, create those silos, but they're all under one pillar, if you would. And having all of that data and how it circulates, really I mean some of it they have to share, right? We have to share as a bureau. So it really gives us a unique lens, Mike. I I know we were having this conversation about which clients share and don't. I really don't have the same challenges that others might face in getting clients to share data because they're usually, you know, obtaining or sharing data uh from us. So we really try to create those bi-directional relationships with all of our clients, whether it be you know on the financial services side, our side, to be able to create those data shares. So we're constantly testing, we're constantly testing to see what the next innovation is, et cetera, et cetera.

SPEAKER_01 8:04

Yeah, Manny, that's interesting because too, like EXL, right? We're not just the back-end collections, right? Like we're making models for marketing. I mean, we use a ton of your data, right? And some gets to marketing purposes, some of it is for modeling purposes. It's throughout the entire lending decisions, it's entire life cycle. Just you know, when you get to like our product like payment, or we don't have access to all of it. We get what who is on the other end and deciding to give us. And we do try to restrict that data because you know, security reasons, everything else. But it's interesting because we've had creditors share, you know, transunion data, and creditors not. And we don't really, you know, what your data, and because exactly what Adam was saying, decay happens, and it happens quick. Like itself into a person who owes a lot of money, and maybe it was a medical reason, right? They didn't even lose their job, but they're behind, or or maybe they lost their job and now they got a new job, but they're still behind and they're trying to catch up. That phone is ringing, it's ringing a lot. They're getting texts, they're gonna they might just say, forget it. I'm getting a new number, I don't care. I'll drop off my friends, my friends can find my new text, I'll get back right. Like it's a pain in the neck, but let's face it, your cell phone is now your digital signature. Like it is not, it's not just a phone anymore. That is, you know, we're looking at the IP address on it. We're looking at all this information on it. So they might just bail out on it to save themselves some pain of fielding phone calls and everything else. So data is people are choosing to deteriorate their data, or it's just, like you said, aging, right? Uh do you see an increase in that because of the size of the debt? Like, do you see like phone number skipping going up or or anything like that, or people shifting their email?

SPEAKER_02 9:56

So, well, Mandy, let me drop something in here because at the Trans Union Summit last year, we saw a presentation where they talked about the average consumer having four to five email addresses. And so that I think is uh that that's a pretty significant challenge there. Like four to five email addresses. When I heard it, I like almost fell out of my chair.

SPEAKER_00 10:15

Yeah, thanks for that, Adam. That's that's that's a great segue. Five email addresses. I personally I have more than that. I have an email address just for my Uber receipts. All of my Uber receipts go to this one email address because I, you know, expense reports, we all deal with that, right? Different behaviors, man. And if I be I'm gonna get on my soapbox for a second, right? But back in 1995, if my behavi my behaviors were limited to what I who I called, who I texted, maybe, and how I paid my bills. That's it. Today, it's who I talk to, how I use my phone, how many emails, how many text messages, how all of it, how I utilize all of that. And the trick is all of that's changed. It works congruently now. I don't, you know, keep trying to tell people stop looking at the channels and look at the mobility. I carry this thing with me everywhere I go. So every channel you use today is being launched at me instantly, right then and there. I'm seeing it. My behavior is telling you something. If I'm not picking up your phone call, it's probably because I don't pick up phone numbers. And you're trying to, you know, trick me with the local anti and stuff like that. And we you should be able to decision that very early in your contact attempts so that you change that behavior, whether it's branding the call, whether it's branding some calls, whether it's using your 800 number. I don't know, but whatever behavior you identify, challenge that problem. And that's where I think e-commerce has really met the customer, right? They follow the breadcrumbs to the to the behavior, not to the solution they want. They'd follow it to the behavior and adapt to the customer's behavior. So think about phone calls. Mike's when's the last time you answered a phone number you didn't recognize?

SPEAKER_01 11:48

I love doing it. So the purpose of the phone call. Adam has a work statement for a lunatic who loves fan calls, normal.

SPEAKER_00 11:57

You're one of the two percent that you're one of the two percent that does. Yeah, you're you're one of the two percent, two point three four percent RPC rates of of people who call. And you obviously probably don't owe anybody any money and you know aren't avoiding any phone calls and you're trying to gain prospects. Or I I have no clue, but you have a strange behavior if you're answering phone numbers because 90 some odd percent of people who are in debt, I would say, because these are collection agency statistics. No, I'm sorry, I'm correcting myself. These aren't collection agency statistics, these are contact statistics from my uh my contact folks. So 90s, I'll say 90 some odd because I don't like to be specific of folks aren't answering phone numbers. You know, so I'm not trying to you know get on my high horse or about branded call display. I'm just trying to say that behaviors overall and the way that we use contact channels, the way that they've changed immensely. They've changed immensely over time, and it makes us more rapid to respond to each one of those types of behaviors. I'm sorry.

SPEAKER_02 12:50

No, but branded call display is something that I wanted to go back to because I I think from a data perspective, one of the objectives that we have as debt collectors is building digital trust. Whether we're sending um MMS messages that have branding, and then we're building our campaigns between MMS and SMS, because SMS was just built to be a notification tool set, not necessarily a branded tool set, but as we're using all of these channels, we can send 10 million calls from an AI voice chat bot out to the portfolio of accounts that we're working. But if we're not leveraging some of these additional digital trust tools like a branded call display, how are we actively getting through? Yeah, I don't answer the phone, right? Three different spam filters on my cell phone to restrict as much as humanly possible.

SPEAKER_01 13:39

I think this is where AI comes in, right? Because as Manny said before, behavior is king to me, right? Like I AI is using everything and it's not siloed, right? It's combining every outreach you can. And even we're using things that aren't used in the US, you know, like not just WhatsApp, but other messenger services all over the world, right? So we they give us a preview of like what RCS is going to be, where branding is a little bit more um available in text messaging. But if you use them together and they're under one centralized intelligence and data is feeding in on hey, ignore this text, but open this email, or the other way around, then I can generate ways. If it's let's say it's a HELOC, right? It's a big balance, it's probably not gonna be negotiated out on a text message, but I can use text and email to drive an inbound call. So now the strategy, you know, I've had agencies or or creditors say, Hey, this is big balance, never gonna work. And I'm like, would you like more engagement? Back to the beginning. The data feeds the engagement tool, which feeds your strategy, and your strategy is gonna change. This is just like it changed five years ago, just like it changed with the telephone. Collections is not second only, it's a very old business, I'll put it like that. And it is not the same as it used to be. And this is it is changing and it's changing very quickly. And the tech is data led. And it is you're using data now to make changes to your strategy, and it's gonna get what I tell clients is phone, email, text together. You're gonna now segment out behaviors to work on behaviors. And the example I always give is I went to the payment portal, I spent three minutes because I'm tracking it, spent three minutes there, left. No payment, nothing, no response, nothing. That is not your typical next day phone call. That is someone who tried, right? Like that behavior is they tried, they made an effort, and that's someone you should concentrate on. And I think we're gonna get when you look at all these different behaviors that are gonna happen and and track now, and you're gonna use AI to track and and help you make decisions, your your strategy is totally different. But it goes back to that data, right? Like you need to be able to reach these people, you need to be able to measure that stuff.

SPEAKER_00 16:08

Well, I think you know, it's too big, right? For if you look at trying to get people to do what you're doing describing, you can imagine the size of staff and how many decisions you're trying to make in one you can't, right? So AI is the most way.

SPEAKER_01 16:21

Yeah. Adam can probably do it overnight. He's channel something, but he's maybe working like a big deck. No, no human beings could do that.

SPEAKER_00 16:30

Like I remember back in in at Green Sky, uh, which was a Goldman Sachs company, we were uh we were trying to figure out uh channel choice, right? And and that I'll just be transparent back then, I'll give FICO a plug. We used uh FICO CCS, and we, you know, for me it was okay, I'm gonna install this tool, it's gonna make all these great decisions, I'm good to go. What I didn't realize was me and my team were gonna go have to go and plug in every single yes-no in that decision tree to give me the outputs and work for and to send things down the workflows results-based, right? And that was like a six-month endeavor of just, you know, pulling my hair out of my head. That's how I got this grave, if you don't know. But it was also one of the most useful exercises I've ever been through because now I understood exactly what every behavior to Mike's point, and that and and I realized this is way too big. If there's a way to really automate this, um, I'd be so much more successful because I can make these decisions in real time and apply them to individual accounts rather than cohorts. And that's what what what what what AI can do for you is it could take all of these types of you know behaviors and say, oh, went to the empty card in the portal, three attempts, three attempts with an Annie, a local Annie, two text messages sent, an email sent, what's the next step? Whatever your decision and your strategy is. And they could decide that for you on multiple scale across, you know, 40 million records at once. And you, you know, and you're gonna pay, I don't know, uh, you know, a minuscule portion of I mean the ROIs are freaking incredible on products, right? I think what was the satisfaction rate? It was um sixty, no, eighty, ninety-eight percent of firms were were reporting some type of level of satisfaction, high or or mid-eye satisfaction with AI adoption and what they've adopted over the last year. It's incredible. It's an incredible opportunity for our industry to really take out, and especially our mid-size, I can't I keep saying this, our mid-sized to smaller partners can now compete on a larger scale because they don't necessarily have to throw a you know 100 FTE and go through all of those these things to build these types of strategies. I'll get off my soapbox.

SPEAKER_02 18:36

No, it makes it make it makes a lot of sense. I think the the challenge here is the application of the data set to the tools, because if again, if we're not feeding it with what it's gonna need to be successful, where are we actually going with the tool set? Now we could use large language models to machine learning opportunities to better predict behaviors and to put the right message or to construct the right content and get that into the hands of the individual consumer. But that coordination and collaboration across channels, that understanding has to feed back to somewhere and it needs to be fed with new information, otherwise, that decay is happening so quickly that we're communicating to the wrong people at the wrong time. And it's hard for us to get that same value out of the activity. It's difficult for us to hire new people. We've seen that across the board. Everybody's struggling with that, and they're really in they're really doing well or they're they're finding satisfaction in their investment in AI. But I think that the view of the investment in artificial intelligence is skewed because they go in, and I'm gonna use it as an example since I'm sitting in the in the middle, at least on my screen. When I go to Mike and I say, okay, how much is it gonna cost for me to get these tools together? And Mike's breaking it down. We're saying it's gonna cost you X amount for this piece or per call or however that particular tool is structured. But I think where the industry is missing is they're not calculating up front. At least those that are not satisfied with their AI investments are not calculating the cost of the data increase that they need to put in play in order for them to feel that satisfaction. So those that are not feeling satisfaction, from what I could see in the data, was a small subset, but that small subset is only looking at it from a technology standpoint. They're not thinking about it as an operator who needs the technology and the data and the marriage of those two worlds in order to create that success or to find that satisfactory.

SPEAKER_00 20:42

If you want to simplify the same thing as a dialing file, you go through a file and you get a 2% contact rate, you see it's bad phone numbers. What do you do? You skip trace the file, you get better phone numbers, and you dial it again, you dial the better phone numbers and you enhance your data. Why? Because the phone numbers were old, they were outdated, and they were no longer you no longer had a good contact rate because you didn't go and enhance your data. That you know, across the spectrum. And you know, it was easier before because we were calling landlines in 1983. Today you're calling you're trying to contact people not just across various channels, but those channels work congruently and they integrate, and you're trying to anticipate who which channel. You could have a great phone number, terrible email address, great phone number. You're trying to figure all and have the best contact channels because they work congruently. If you want to reach me, you're likely gonna have to make different types of and if we're talking specifically about contact, it's likely gonna have to make different attempts to try and reach me. And if you don't inha have the right the right contact information for all of those channels, you're missing a piece of that strategy which is going to entice me to either answer your contact, reply. And like I said, the purpose has changed. I ask operators all the time, what's the purpose of a phone call? And they told me to get for an RPC. And I tell them, No, that used to be the purpose of the phone call. Today the purpose of the phone call is the same purpose of every communication channel payment, the result. That's the purpose. You don't necessarily have to talk to people to make you get payment anymore. But how you know that and how quickly you can get to that determines your margin, which today is pressed more than ever, with the higher volumes agencies are facing, the lower, lower liquidation rates and the margins getting squeezed. AI is a real solution for those types of problems.

SPEAKER_01 22:17

I think too, many, the other thing you're you're doing on that phone call is asking them how they want to be communicated, right? Like part of what AI does well is that e-commerce, Adam, right? Like the how do you want to do this, right? Do you want to do it by text? Do you want to do it by like it is just a tool? Phone is just a tool. It's just one way to do it. It is amazing to me though, like when getting back to data is and what I was going to jump in with is we have a lot of clients and we provide a ton of behavioral data back, right? And they don't use it. They can't intake it. And I was like, that's the problem. Right. And I think some people build or buy a tool and they're like, this tool's great. I get this much more contact rate. And I'm like, that's great. But you could do way better than this, right? And and like we're now have like I think we've gone through the first rung of AI where people have tried it and yeah, it's better than no AI. And now we're getting into why is your AI better? Show me, right? And and it's it's interesting, it's changed over the two years I've been or two and a half years I've been out of Excel. First, it was trust me, it's gonna work, it's okay. You know, next is wow, it does work. Third is how is it getting better? Fourth now is how is it better than everybody else, right? Because let's face it, the market is flooded with with tons of things. And part of the thing I think now is that not only the data you put into it, but the data you get back and how that changes and how you can then change your whole process.

SPEAKER_02 23:50

It's the same as our other communication channels. When we think about it like this, it is it is an evolution. But when we talk about we used to say text message SMS as if it was just straight up interchangeable because there was a time in which it was. And then the introduction of MMS and how does that have an impact? And I think that's what we're seeing from an AI perspective now. At first, it was let me set up the train tracks that allow me to send a text message. And then it became, well, how fancy of a train can I put, or what order should my cars be in in order to maximize the load that I can move down this train track? And I think we're looking at artificial intelligence in very much the same way. But if we're not, what are we sending down the tracks when it comes to artificial intelligence? We build the train tracks. And now what are we sending down it? We're sending data, right? The data's the navigator there because the train tracks aren't even going to the if the train tracks aren't going to the right location, if it's not a straight shot from New York to Chicago, are we really accomplishing our goal here? That's where I think the the lack of of or really the next phase of our evolution is gonna go. If we if we I agree and doesn't have text messaging, okay, let's hear it.

SPEAKER_01 25:02

Like you're right. That is, but I think now too, it's the data we get back and what do we use use it for? How do we and I and I think I've talked to Manny about this. I I I think part of what's gonna happen is you know, not everybody has a data analytics team like you know, excel does, right? There's a the there they're or a data scientist that can say, okay, we need to look at this, this, and this, because this could affect our bottom line. But I think the tools for that are being built. That is the next, not just of what you gotten and what you've increased productivity and efficiency. Now, how do you push it up? How do you take advantage? Because, you know, um, this is everybody's putting AI into their arsenal, but are you putting it into an armored car or you put it into a tank? Right? Like you're gonna have to build a better, you know, like you're gonna have to build a better machine. Um, and that I think is is where we're gonna go from five years from now. Like, is that is okay, I've done all this, I'm getting this data back, and here's how we're using it. And I think our RFPs are gonna say, you're using AI, what's it capable of, what is it doing, and how are you using the behavioral data that comes back?

SPEAKER_02 26:12

How are you learning with it? How are you using it to learn and improve your future processes?

SPEAKER_01 26:17

How are you segmenting? How are you strategizing, right? Like it gives you, as Manny said, what he was trying to do at Green Sky in ten years ago. It's it was a perfect idea, right? It's what AI does, but it does it so fast and at such scale. You don't need 40 data analysts to look at it, but you probably need two, right?

SPEAKER_00 26:36

Yeah, look, I think um I talk to Adam all the time, Mike all the time, and I'm you know, I'm a data geek. I can sit here and geek out about information. So I work in the best place in the world, right? I mean, I work at TransUnion right. I have so much information at my fingertips, it's not funny. Even things like I I you know, I I'll geek tell you honestly, I geek out with my buddy over in uh gaming, uh, you know, table games and Vegas type of gambling stuff. Um I geek out with him in the way that we utilize that data. I just, you know, I'm I'm a data guy. So, or I'm a data geek. So when you look at the collections industry and what we're trying to solve, what we're trying to solve as an industry for our companies and our strategies are, and I could provide all of this data, right? How do I reach somebody? But not just how, where, what's the best time, what's the best phone number? This is all setting us up for the most likely success from that communication, whether it be an RPC or a payment, right? Because you're launching this communication to the best phone number, let's say, at the best time of day, because you know, we have all of this very unique information with our acquisition of New Star, Factor Trust, various other companies. We have all of this very unique data that tells us when the best time and best information to contact somebody is. I could tell you that we want to know how likely we want to contact an order of the highest likely to pay. Who's who has the highest propensity to pay their bill based on their financial behavior or financial information, whether it be based on current information, you know, we have credit bureau information, bank information, you know, all the FCRA permissible stuff, or um, if it's older accounts, usually, or sometimes in current industries, triggers, you know, event-based triggers. They might get a new job, they might buy a new car, they might do, you know, all of these, you know, we I could provide you with all of that information as well. I can tell you, you know, I'm working on this is a really cool new product that we're coming out with and I could share now. Address behavioral intelligence, digital signals at an address that tell me, based on your digital signatures, whether Mike is ordering his DoorDash. Well, I know whether Mike is ordering his DoorDash at his house or at his buddy's house. And based on frequency, timing, all of that, I could assert that's an active address for Mike Walsh, right? And I could give you all of that information. And I could probably give you, I could sit here and I could go down my product line and you know make my bosses very happy by talking to you about all the capabilities that TransUnion has. But instead, I'll tell you that being able to put all of that together, I made a career out of doing that stuff. I think many of us here have at some point in our careers, right? Now, today, you know, I'm not saying that I'm not I'm not trying to insinuate, you know, this the Skynet takeover or anything, but but you could do a lot, so much more, so much more effectively and decisioning, so much more compliantly. Imagine if you could, you know, back in the day, if I could just plug compliance into all of my talk-offs, you know, and all of my outreaches. And today you can do that automatically if I could just, you know, and have zero risk. And that decision is what I told it to do. You ever have that frustration? They're not doing what I told them to do. Well, you could just get rid of that. All of it. I'm just saying, I've been in, like when I worked at at eBay especially, uh, you know, being in that, I'll call it, Adam, I'm gonna use your phrase, the that I was at that intersection of e-commerce and collections, trying to collect it eBay. You know, and having everybody on that one platform and being able to have all of those behaviors and seeing the way that e-commerce looked at things versus the way that I looked at things really had an impact on my career. And I think, you know, it's been obvious since then. Um and look, I remember a guy, I I I'm gonna give him another plug, Dr. James Ward helping us with the the, I want to say, deploy artificial intelligence in the EU with integrity and compliance, which was very difficult for us to do. We we were doing it across the whole company. So think, you know, the way that people search for products, the way people list for products, photo recognition, things like that. My play was the decisioning that we were putting into our casual civil management programs and how that was working. And and working through all of that, what you know, that's what led me to I can do so much more with so much little and be so much more accurate and effective. And not necessarily, you know, I can if I want, I could be an FTE play. You know, I'm just gonna be real. If I want to reduce FTE, I could probably reduce FTE. But what I'm really trying to do is meet the customer where they want to be met. Because I don't want to spend half an hour with your agent having to explain my life situation. I want to go and pay you on a portal, or I want to pay you the portion I want to pay you, and one I want to pay you, and I want to be able to negotiate that. And the quicker you can get me to do that, the more successful you are at an agency. And decisioning that is something that I don't do. I don't provide those decisions for you. People like, like, you know, that's what I love about Mike's partnership with uh with TransUnion, because we can supply him with so much information that, you know, the layman and the agency either can't get or quite frankly can't afford in their margins, but you can put together strategies that most people can and help them put together these strategies. It's just a finite level, it'll be so efficient and effective that it just not only betters our industry, but betters bottom lines. And I love that.

SPEAKER_02 31:15

So I want to go back to something, Manny, that you said, because as you were talking through kind of this change over time and and how things have changed, but I I pulled up one of the statistics from the Transunion Debt Collection Industry Report that we just released. And I what I thought was really interesting is there was a 10% increase between 24 and 25 in the groups that responded that their AI investments were exceeding their expectations. And I don't think that that was I don't believe that that large differential was driven just on the change in the tools itself. I think companies are getting smarter at how they're actively deploying it. And if they, like I was saying before, if they start looking at artificial intelligence investment and they start lumping in some of that data change or increase in data or how those data strategies into it, I think it changes the game for them quite significantly. Like that's where the that's where the gold is hiding in them their hills.

SPEAKER_00 32:10

There's so many different companies out there that can help us support you. I mean, like like Mike's. What is it? Less than another statistic that I'll quote from that report is less than 10% were developing in-house. Mm-hmm.

unknown 32:22

Yeah.

SPEAKER_00 32:23

And almost no creditors. Right, right. Creditors, especially on that front, they were the ones really going out and trying to buy from companies because there's so many companies out there, and there's so many people offering different AI tools for you to utilize. And I think it's very important. Last soapbox here, I see, and I think I've expressed this to Adam and other folks. I don't think I've talked to you about it, Mike, but look, try to solve your problem. This is my advice to the to the industry. Look at your problem and think about how you can automate it, even if it seems impossible, but think about how you could automate it, but solve a problem. Don't try to just implement some type of AI products or you could put it on some type of sales literature and tell your clients you have AI. It won't work. I'm seeing it for the sales literature anyway. And you know what? If it's if it's AI, please please learn the definition and make sure to talk to somebody in data, make sure that you're that it's not just automation, because you could automate a lot. Um, but there has to be some learning involved. There has to be some large manipulation of data, there has to be decisions that you've actually given away, which are rather challenging to do for folks. I mean, I know if you told me I'm giving away decisioning on how to communicate somebody and I'm not in those dialer meetings every month like I used to be, I'm gonna tell you you're crazy, right? But today, if you could get rid of those decisions and trust that those decisions are being made exactly how you said they were, you're just gonna be a lot more successful. But AI itself, yeah, yeah, learn the definitions and don't just try to join the bandwagon here. Try to solve your problems with these tools, and you'd be surprised how successful they are.

SPEAKER_01 33:51

Man, it's it's funny because we haven't talked about this. And the first thing we start with is what outcome are you trying to achieve? Right? Like, start with the outcome. Best question. And then it is that's what we need to solve for. That's what we that's what AI can solve for. If it's an outcome that is based on other external factors, you know, like especially when you get to customer service and other things like that, but maybe we can't, right? Like that, but that's where you start. Collections is it's just customer service, it's just really hard customer service. So that's what I felt that yeah.

SPEAKER_00 34:21

Look, you just hit the nail on the head. It evolved. Collections has evolved and our tools have have evolved over time, but the utilization of our tools have evolved over time. But our decisioning or the tools to decision, the tools to work large sets of data, you know. I mean, look at some of these inventories that people are trying to manage, man. It's it's incredible. And behaviors are now, again, like we said, they were limited to one channel. So everybody kind of behaved the same, right? Seven o'clock, everybody was kind of eating dinner. Time time, remember? And then then came the advents of cell phones, remember? Let's try to catch them on the drive home because that's really when you handled your business now. You didn't do it at the dinner table anymore. And then then we then we started adapting text messages to that's point. Text message and SMS to different things, tended to be a notification, but we decided we're gonna try to collect debt with this. We're gonna write books.

SPEAKER_02 35:10

You're gonna write books in 32 cars.

SPEAKER_00 35:13

Yeah, and then regulation wanted to take up half our books, right?

SPEAKER_02 35:16

Yeah.

SPEAKER_00 35:17

And don't get me wrong, somewhere in between, we decided that email is now a viable contact tool and started using that instead of fax machines and and and things like that, right? So as we've seen this industry evolve and the tools evolve, AI is just the next evolution, right? There's more debtors than there's ever been. Don't get me on my soapbox about the consumer situation. But you know, just look at the portfolios themselves are larger. Look at student loans. Whatever that bubble pops, I mean 1.4 billion consumers? Hello? How are you gonna work all that debt? You're gonna do throw throw half the country at it? You're gonna hire half the country in your call center? No, man. You're gonna have to figure out ways to, you know, rag models or regulate all that data, get through it, uh and some ideas on how they work. If you're an ACC today and you're not looking at these tools, I I I love the where was the line? Uh, AI has gotten mainstream. It's only 7% of companies in 2025 saying they don't intend to deploy AI or ML tools, um, which is down from 27% in 24. That's a 74% um year-over-year decline in the non-adopters, and that's from the debt collections industry report. That means you're in the 7% of people that aren't gonna have these, I won't say this, but um these advantage-giving tools that are gonna give your competitors more speed, more accuracy, and the ability to be able to create strategies that we used to dream of only two, three years ago.

SPEAKER_01 36:41

And one more thing, many, scalability, right?

SPEAKER_02 36:43

Yeah.

SPEAKER_01 36:44

Your competition could scale from I literally this is a real world result, 30,000 accounts to 350,000 accounts overnight. You know, like overnight. Not worried, just going.

SPEAKER_00 36:56

And you could collect payments at three o'clock in the morning. Correct. 24-7 on a Sunday. I love the idea of scalability, Mike, because I truly believe that you know, when if you remember when AI was first introduced, it was you know, it was expensive to buy a large language model. I mean, you it's cost prohibitive. There's obviously with competitors popping into the marketplace, that's driven prices down. And now, if you want to implement the large language model and you have the team to be able to do that and know how it the gumpture will go ahead and do it, you can probably afford the tool itself. The thing else is still a little more challenging, but you can probably afford the large language model itself. I think if that continues to grow and adoption continues to increase, and we see competitive pricing kind of normalize it and meet the market where it needs to be, it's continuing to give our smaller and mid-sized partners the opportunity to compete for larger um RFPs and bids that they wouldn't be able to bid for before. Being able to work huge account sets now with, you know, their 40 and 50 FTE not requiring, you know, a thousand or 150 dedicated FTE. You know, they could work larger volumes of accounts, you could be smarter about the way you're you work your accounts, you could beat your competitors on scorecards, all because you're buying better tools. You know, give me a give me a hammer, uh a physical hammer, give me the staple hammer, I'm gonna beat the guy with the automatic hammer every time, right? Vice versa. I don't know. You could tell I'm not a construction guy, but you know what I mean.

SPEAKER_02 38:18

So I I have one more thing that I wanted to throw out there for my outline before we wrap up. And I think it's the risks associated with leveraging these tools at scale without improving your data strategy. And one of the things that I've heard through a couple of podcasts that I've done recently was, for example, when we talk about text messaging and we talk about how tightly controlled, for example, RCS is going to be as it even becomes available to the debt collection industry. When it comes to email, when it comes to the text messaging, like if we're not putting the right numbers behind it, we're gonna start having more carrier problems. Because the more email, the more bad emails that I send out, the more spam notifications, the more bounce backs I get, and the less likely Gmail or Yahoo or any of these other more common consumer email systems is to actually deliver my messages.

SPEAKER_00 39:09

All right, I'll jump in. True. Uh one of the I I I one of the ways that I was even introduced to TransUnion, one of the one of the acquisitions that I think was most impactful to TransUnion was the acquisition of New Star, which I think happened about four years ago. Uh the integration of those two companies and the type of data they provide was immense. When you think of TransUnion, you think of uh Credit Bureau, right? I don't know about you, but when I think about credit bureaus, I think about old, fussy, you know, musty places. And transunions did the polar opposite. That's why I love working here. So New Star has, you know, very unique contracts with service providers. They owned um caller ID way back in the 70s. They still, you know, they owned that trademark, and because of that, had very unique relationships with all the carriers, and they've been able to create propensity modeling and scoring based on best time to call, best number to call, all of that in combining the same. Those types of acquisitions and contact opportunities have always given us the opportunity to expand into other areas and be a little more proactive and forward-thinking in what we do and how we do it. How you contact people, when you contact people, and how you engage people is really really the heart life of our industry, right?

SPEAKER_02 40:18

So that was the piece that I wanted to hit was it's the the risks that we The risk. The risks that we run in terms of actually sending these outbound messages to the wrong people is not just a regulatory issue. It is literally the carriers making those decisions. ATT, Verizon, if we're sending bad text messages, they're not going to let us send text messages anymore. It's a much more tightly controlled animal when it comes to the text messaging channels. And I think when it comes to the emails, it's still a risk.

SPEAKER_00 40:47

It's trusted solutions there that can give you that. There's ways through the carriers, and I'll explain it this way in the most layman term I can think of. There's ways that you can create your blue check like you do on social media to certify that that's you, either through caller name optimization on phone numbers, but we call them our trusted cost solutions subset. You can the the way to combat that, because to your point, as an industry, we have to be careful with our behaviors, and I think we've been good stewards of communication channels until now, whether it be through regulatory enforcement or just actually consumer behaviors and expectations, and we're us learning to meet customers where we need to meet them. Um but the risks we run are severe, right? And it's not just contacting the wrong person, but again, it's also wasting time and money doing so, right? And being able to make sure that you're sending the right message to the right email address, the right phone number. It's just become table stakes at this point. It's not like it used to be where this was a fancy offering. I think that's why we probably didn't bring it up till last minute here, because it's table stakes. You need to be taking those type of steps just to protect yourself um and and be fair to your customer, your c or the consumer you're trying to reach, because in today's age you have to be.

SPEAKER_01 41:53

I think too, Adam, you you can use AI as like there is an art and science to sending email text messages, right? Again, it's huge data sets. It's it has to be monitored in real time. Like, yeah, you can send you know 20,000 texts out and hope for the best. Or you can send them out one per minute and monitor what's happening, right? And monitor, wait a minute, is this file full of garbage? You know, and we should stop this campaign and and redo it. Like that's what these tools give to you, right? You can monitor spam filters and like if you're not doing this, your deliverability rate, you're killing yourself, right? You're not getting the engagement out to the people you're trying to engage. Manny said it earlier, like this data helps your engagement tool, but your process should be now it's table stakes. You should be using the best tools you can to reach customers because frankly, your competition is doing it. The creditors you're getting accounts from if you're an agency are doing it. Like it is had an agency manager say we are now software managers. We heard software companies.

SPEAKER_00 43:01

I think it's both, right? Your process is only as good as the data you put into it. And that is only as good as the process.

SPEAKER_02 43:07

Yeah, I agree. Gentlemen, I can't thank you enough for coming on and sharing your insights with me today. I think we've done a great job of talking not just about the theoretics of what is artificial intelligence and how is it going to impact our business, but talking about how it's actually being applied to the space and how we can improve those applications. So thank you everybody for listening to Applying AI, where we explore how to make artificial intelligence work in the real world of regulated industries. Thanks for having me. Subscribe to the show, subscribe to the show on your favorite platform or YouTube, and find more insights and resources at receivablesinfo.com. But thank you, everybody, and we'll see you again next time. Thanks, guys. I appreciate you.

SPEAKER_01 43:46

See ya.