AI Use Cases in Debt Collection with Latitude Software | Ep. 7

Adam Parks (00:07)

Hello everybody, Adam Parks here with another episode of the AI Hub. And this is one of my favorite things to talk about artificial intelligence as it relates to the debt collection industry. And when we first started the AI Hub series to kick it off, Cris from Latitude and I did a live webinar episode talking about the six use cases of artificial intelligence for the debt collection industry.

And that was about two years ago. want to say that was December about two years ago now. And I learned a lot from that discussion when we started thinking our way through what are the actual use cases because debt collection and businesses in general all have different use cases. We could talk about accounting and HR and all of these things that apply to everybody. But for today’s conversation, I really wanted to hone in on talking about those six use cases that are directly related to the debt collection industry.

Adam Parks (01:03)

and no one better to have that discussion with than Cris. So Cris, thank you so much for coming on, sharing your insights with me today. I’m excited that we get to continue this conversation.

Cris Bjelajac (01:13)

Thank you for having me and thank you for the kind words. I appreciate it.

Adam Parks (01:16)

Well, Cris, I know that you’ve gone through, you’ve gone through some transitions recently as an organization. And for me, think Latitude has been the perfect partner for this particular series because you’re to me are the AI hub, right? You’re the spoke area, the center of the wheel and the spokes of all the different AI use cases can plug into your platform. But could you tell for anyone who has not seen previous episodes with you before, could we start by you telling a little bit about yourself and how you get to the seat that you’re in today?

Cris Bjelajac (01:46)

Yeah, sure thing. I’ve been working with the Latitude product since 2018. I go back in this industry back to the early aughts. I worked for a company called SoundBite Communications, an outbound dialer, SMS and email platform. And then when Genesys acquired SoundBite, I was asked to look after the Latitude line of business, which was also owned by Genesys at the time back in 2018. For those of there may be some people are not aware Genesys Decided to sell that line of business to TEC services group. So I’ve just came on board to TEC Services Group about two months, exactly two months ago and thrilled. It’s going great. And so we’re in the process of transitioning the Latitude line of business, figuring out our investment strategy going forward. You know, still a lot of activity, you know, in that arena, lots of interests on buying Latitude. So as you know, new customers are coming on board. it’s just fast, fascinating, fun and exciting and just everything’s moving super fast.

Adam Parks (02:49)

Well, very exciting. And it’s not like the TEC coming in and starting to work with Latitude in a new way was that different because you had such a deep partnership between the organizations leading up to this transition.

Cris Bjelajac (02:56)

You know. Yeah, for those people who are not aware, TEC was our North American reseller partner, exclusive. so from a day-to-day operational aspect, not a huge change for me. I just spend more time with the guys at TEC, which has been great.

Adam Parks (03:20)

So for anyone who is not that familiar with Latitude Software, which I would be pretty surprised because this platform has been around longer than I have in the industry, take just a minute here and see a word from our sponsor, Latitude Software.

All right, so now that we’re back, the first thing that I wanted to talk about, Cris, was I thought that we should frame the conversation around the six use cases, and then we can go through, I’d to talk a little about how different types of companies, debt buyers, agencies, law firms, creditors are leveraging these tools specific to the use cases of their business model. And then we can talk a little bit more about each one of the use cases.

But first things first, right? The six use cases, the way that I see them, and I’ve had this argument both on stage throughout the years now and also on a variety of podcasts and conversations, but I look at it as these six. Quality and compliance monitoring, chat and written communication, scoring and treatment strategy optimization, voice AI and real-time agent support.

Adam Parks (05:06)

negotiation support and offer modeling. So that whole negotiation piece and understanding that from a large data set. And then operational analytics and risk forecasting, which I think each one of these kind of fits into different boats. But any piece of technology that anyone has showed me across the entire debt collection industry, I can put it into one of these six buckets fairly comfortable.

Cris Bjelajac (05:23)

Yeah.

Adam Parks (05:31)

I haven’t seen, I mean, there’s some organizations that do many, but have you seen anything that like really falls outside of those six?

Cris Bjelajac (05:39)

No, outside, know, specific to the collections industry, no, are they using other AI tools to help the, you know, the back office of the business, you know, stuff that every business might be using? That, you know, that’s definitely happening as well. But for the collections industry, I think those six buckets certainly cover what is being used and being purchased today.

Adam Parks (05:59)

Absolutely. even looking at everything that I saw, I mean, there was a ton of AI vendors at the Debt Connection Symposium. There was a lot of AI vendors that were at the ACA Annual Conference. I heard, although I wasn’t there last year, a lot at the RMAI conference. And just looking at the sponsor list for 2026 event, I can see that there’s quite a few more that are going to start coming to the forefront. It feels like a lot of the new progress has been directly related to voice AI.

And it makes me a little bit nervous because I feel like voice AI is one of the more, I could see the benefit obviously of both managing inbound or even making outbound calls using that technology. But it seems like it comes with the highest risk level. It feels like there’s some low hanging fruit here in terms of being able to do scoring and treatment or quality and compliance monitoring. Like those seem like the, again, more back office, less consumer facing, which

Cris Bjelajac (06:53)

Yes. Well, I looked at it way. I think the voice AI, the agent assist is something that agencies need to, don’t have the in-house expertise and they must buy it from the marketplace. Whereas I think scoring and quality that…

Adam Parks (06:54)

comes with less risk.

Cris Bjelajac (07:12)

you can with a good account, know, with a decent sized account on ChatGPT or, know, your favorite engine, you can build a lot of that in-house. And we are hearing that from some of our, some of our customers is that.

They just create their own ChatGPT product. It might be some piece of software that they vibe coded, you know, and is running on their own systems or they simply have a process to run it through an open ChatGPT that is cordoned off with just their data. But when you start talking about, you know, agent assist and things that, you know, really needs a high level of technology, you know, the connectivity to the back engine and the compute that is required to use those types of products, they’re going out to the marketplace. And I will say, the difference between even six months ago of our customers talking to today, Everybody’s on the agent side of things right now. They’re all looking at the various players. Six months ago, I might have been like, you can hold off, take a look, see how the industry shakes out. Now, everybody’s really assessing where they’re going to spend their dollars in 2026.

Adam Parks (08:11)

I think that’s a pretty astute observation. You know, for me, it’s like the voice AI that’s completely autonomous in communicating with a consumer versus the real time agent assist. I’m more comfortable with the agent assist at this point because that human still remains in the loop. And that’s not to say that I haven’t done some really great demos, but I’m really careful about how I allow a demo to happen. Like if you’re just going to play it from me, like if you’re just going to play this call from your phone and have me listen to a pre-recorded, I’m not all that impressed because I can make my computer say anything that I want. I have time to manage that. I have a lot more respect for the organizations that will put me on the phone with it and let me have a conversation directly with the bot itself. That’s always the case, but especially when it comes to the voice stuff, I always ask the first same two questions of every organization that pitches me on Voice AI. Who’s your lawyer?

Cris Bjelajac (09:10)

you

Adam Parks (09:20)

and who’s running compliance. Those are always my first two questions because I want to know a lot of these organizations are coming into the United States. They’re operating in debt collection for the first time. And let’s be perfectly frank, it’s one of the most regulated industries on the face of this earth in one of the most regulated economies on the planet. So.

Cris Bjelajac (09:30)

Yes.

Adam Parks (09:39)

It’s important that we are looking at the underlying tool sets and understanding the privacy implications of all of this, because not only if we’re taking payments through this channel, you know, we’re talking about, have PCI DSS and from what I’ve seen, you know, adoption of the conversational AI tools has been the lowest of all of the, all the use cases that I’ve seen so far, at least from the survey results that I’ve seen. from 24 and 25 for the TransUnion Debt Collection Industry Survey. So, and I want to say it maxed out around, those that deployed it, that was only representing roughly between one and 20 % of total collection. So nobody seems to have necessarily cracked the nut at scale to be able to deploy a tool like that.

Cris Bjelajac (10:26)

No, but I think the next survey is going to be entirely different. if chatter inbound to us, security assessments that are sent to us to provide on behalf of our customers is any indication this has gone from, you know, it was exploratory and now all of I, you know, that’s a bold statement. Maybe not all, but most of our customers are being asked by their customers, the banks, what are you doing? I want to know what your strategy is. So a year ago, the questions were like, we’re not really sure. I want to see what you guys are thinking about and look for risk. They were looking for more risk. Now they’re looking at their vendors to be

Cris Bjelajac (11:15)

bringing to them awesome AI solutions and better operational metrics that they can get better results from their agencies and have them competing against each other. So my recommendation to not only our customers, everybody in the business is if you’re not checking these things out, you need to be doing it right now because I can guarantee you, your competitors are right now.

So I think the debt collection industry is conservative. I mean, I lived through the whole text and I work for a company that first wants to do text messaging and collections. And it literally has taken a decade to get to there. That is not happening in AI. That is…

Adam Parks (11:50)

Very.

Cris Bjelajac (12:04)

It is accelerating as I think as fast as it possibly can right now. As fast as this industry is able to absorb it. I don’t see any, I don’t see any, I don’t see any regulatory roadblocks. Everybody’s like, how can we use this? How can we go and keep moving? I mean, I’m not, I’m just.

Adam Parks (12:24)

So we just saw it yesterday. Trump is expected to sign an executive order in short order here to establish a federal guideline for artificial intelligence to preempt all of the efforts of the states, which I think is one of the smartest things we can do as a nation. Because if we end up with a patchwork of states passing rules on things that they don’t understand, we’re going to create an anti-competitive environment for the United States on a global scale.

Cris Bjelajac (12:52)

Yeah, I, yes, I think everybody’s looking at those regulatory headwinds and saying, is different. This feels different. This looks different. It’s, yeah. And if that is successful in impeding the states from getting involved, then that’ll be.

radically different from then all the other technologies that we’ve all been using over the last decade, including tech.

Adam Parks (13:16)

Well, privacy has been the issue in the past, right? We have these privacy issues, have privacy rules that are coming up in all these different states and it’s almost impossible to comply with all of them simultaneously. And we’re spending more time and money on managing bureaucracy than actually innovating, building technology to drive our businesses. So hopefully within that framework, we can suck in some of this privacy and establish that because I think some of the rules that get passed by states are passed by individuals that don’t have an understanding of what they’re. And if they don’t have an understanding of the underlying technology, how are they going to be the ones to set the rules?

Cris Bjelajac (13:46)

Yeah, and.Correct. then, you know, I also, I think, you know, because of all that regulatory infrastructure that we’ve all had to endure over the last 10, 15 years. You know, you’ve already, everything’s kind of gotten settled in my opinion. know, like all the, you know, there’s no new sort of thing. Applying of AI sort of fits in with everything else, right? It’s not like SMS, you know, like a new channel or something like that. It doesn’t Yes, is it revolutionary? it going to, you know, it’s an amazing technology. Is it extremely different? But it still operates within all the parameters that we have. We still take payments through the credit card system. We still have websites that will drive debtors to make those payments or we’ll take them over the phone. From that perspective, the regulatory framework around it is not that big a difference. This is about operational efficiency at the agencies and debt buyers and all those types of things it becomes extremely valuable. And those are the conversations that I’m having with our customers. They’re like, hey, I want to integrate this new AI tool into Latitude right now. Like now, not 16 seconds now. Like now.

Adam Parks (15:12)

Like yesterday, but having a system built on open API is what makes that possible, right? Like that’s, you’ve already given them the rails by which to do it. And you bring up something interesting that things are remaining the same when it comes to the payment technology, right? Like credit card is the number one and how that starts to function. What I find to be interesting about that is I’m recording this episode. I’m sitting here in Campo Grande, Brazil. And what I’ve learned in the last couple of months here is like they have a, they have a payment system that’s built into their banks. It’s called PIX. It’s similar to the UPI platform that’s used in India, where if I want to go buy a coconut water down the street, I pull up my phone, I scan the QR code, I pay the guy, right? So that platform exceeded credit card transactions in this country this year. So that platform has done significantly more in terms of payment. volume and number of transactions than credit cards have done within the borders of this country. The same thing is getting close in India. So I’m starting to say like, when we start looking at artificial intelligence, how many other things are going to start to get disrupted? Is it going to be MasterCard Visa in American Express in charge of all the payments in the future? Or are we going to start to see these other opportunities bubble to the surface so that we can engage with consumers in new and interesting ways? I don’t know if that means crypto or whatever, because I feel like that’s got a much longer adoption curve. But I think it’s interesting to start thinking about some of these things. And, you know, we didn’t really think about wearable technology five years ago. Now everybody’s got meta glasses. Like the world has just changed for us. And I’m wondering how the application of that technology, because we weren’t texting 20 years ago, right? Like texting took time, took smartphones, right? The 2007 rollout of the iPhone kind of changed things. I mean, you could argue that BlackBerry was already on that. That’s a little bit of

Cris Bjelajac (17:02)

Yeah, I think it’s important that for all these software companies that are building the AI tools is they’re caught in different sort of situation. They need to hype AI as the coolest new amazing technology. You need those guys, the owner or the guys who run the various.

AI companies, the guy who runs ChatGPT, his name escapes me at the moment, but they need to be evangelists and sort of, yes, they need to be evangelists and sort of hyping that this is cool, this is the new technology because Wall Street needs it and all those types of things. getting back into it, if you look at AI in a small business or a medium to any size business, it’s really, it’s…

Adam Parks (17:28)

Altman.

Cris Bjelajac (17:45)

It’s not doing anything. It’s adding efficiency when it comes to humans. And this is the negative thing about it. somebody asked me, my dad, asked me to explain AI, because he’s old and he doesn’t understand any of it. I said, he’s like, so it’s like, can think. And it’s like, well, it can’t really think yet or anything like that. But if I had to describe it, dad, it’s like we have an intern at the company. who just, who’s like right in the middle of his business school, you know, he’s on the summer break and he’s interning for us. He’s incredibly smart. He’s incredibly energetic. He’s always asking what he can do for you and that type of thing. But, you know, he doesn’t have a lot of seasoning yet. And that’s sort of how I describe, you know, that.

Adam Parks (18:30)

That’s a great explanation. Now I have realized that like Perplexity gives me a lot less hallucinations and accurate links for research and is better at language translation. And then I use Claude for other things. But if you have the paid version of Gemini, Claude, ChatGPT and Perplexity and you ask each of them, and you say, have all four of you, I have access to the highest level of all of your models. How should I use you together? It will give you all of the models seem to know where they fit.

Cris Bjelajac (18:34)

Yes.

Yeah.

Adam Parks (19:01)

within the ecosystem of leveraging them for business purposes, which I thought was interesting because I all expected them each to be like, I’m the best. ChatGPT was the most confident. The other ones were clearly ready to defer and say that there’s a better way than just using me. You can use me in conjunction with these other things and Perplexity for research and then organize the ideas in ChatGPT, refine the language. absolutely.

Cris Bjelajac (19:25)

So you have to count on all four?

Yeah.

Adam Parks (19:29)

Cris,

I carry a secondary phone. This is my AI device. only use it for, like, that’s its entire purpose in life is to use all of the different AI tools with each other so that I can start playing with them. Like my wife and I went to Argentina for Thanksgiving and I don’t know anything about Argentinian wine. Now I know a lot about Argentinian wine. I built a model. I figured out how to take pictures of a wine list and have it selected based on my taste profiles and really just started getting into playing with it.

I wrote a storybook for my daughter using one model and then illustrated it using Gemini because Nano Banana just got integrated into all of the Adobe tools. So I think that one’s better for graphic and refinement and Claude is better for coding and final refinement of ideas. But I think each one of them has a place to play. bringing it back to the debt collection space specifically, as I went through…

I use some AI models as I analyzed all the data from the surveys, all the data that I had for the last couple of years before I wrote the 2025 report. And what I thought was interesting was how each company type is leveraging artificial intelligence to match their own business models. So I’ve got a couple here that I wanted to just kind of organize in my head here real quick and talk to you about, because I think there’s some interesting approaches here.

So debt buyers, for example, are focused on analytics scoring models and liquidation forecasting. So they fall into really one to two of the use cases when we start breaking it down from the six. But what I thought was really interesting is that debt buyers were the most likely to want to build AI in-house. And I think it’s because they look at it as a…

competitive advantage and they don’t necessarily want to just buy something out of the box because if 10 debt buyers are using the same analytical and scoring tool and they’re coming to the same bid, they have no competitive advantage. So I think from the debt buyers point of view, the ability to score and understand and predict the liquidation forecasting allows them to set pricing and that’s where they look at artificial intelligence as an opportunity to create a competitive advantage.

Cris Bjelajac (21:43)

Yeah, I would be hard pressed to find, to buy that on the open marketplace. Even though if a vendor came to me and showed me some really great results, I don’t think that the…

think there’s a great advantage. I think that’s something that you can build in-house if you have that. Maybe you’re not technically savvy with in-house, but I think you can find somebody to get to you there. So if you were buying that sort of operational intelligence from a vendor on a subscription basis, I…

about a year from now, you might say, am I really getting the value out of this that I couldn’t just get myself on? So those are the things that, know, that’s build versus buy. I sell software, I’m typically a buy guy, but you know, I think some things you can build in-house with these tools. And so it just takes a bit, slight leap of faith to sort of give it a whirl.

and assessed, know, all right, I just got a portfolio in, let’s run it through, you know, what I think is the right system and let’s compare it to the tools that I’ve used in the past. I think you’re going to get to that and exceed that rapidly. yeah, that would not be an area that I would recommend, you know, to any agency owner as a buy right off the bat. I think that’s something you can build.

Adam Parks (23:01)

I think that’s a hyper focus to the debt buyer situation. And if you look at the earliest companies that came to the debt collection marketplace, I want to say some as far back as 10 years ago, who were talking about the deployment of artificial intelligence, this is where they were deploying it. But I don’t see any of those companies anymore. know, Atunely and there were some others that were selling similar scoring mechanisms that I just I haven’t seen them in a long time. Yeah, I don’t think they’re still in business. But I think that

Cris Bjelajac (23:20)

Yeah. Yes, they’re not going around. Yeah.

Adam Parks (23:27)

kind of points out that like it really is one of those things that’s proprietary. Now, as we move on to the debt collection agencies, right, which is more of a customer experience type of business model and organization, the emphasis is on digital communications and QA automation. So being able to understand the quality assurance and be able to do that at scale. There was a time where we listened to 5 % of the calls. Now,

Cris Bjelajac (23:54)

Correct.

Adam Parks (23:55)

The AI can listen to 100 % and the same people that we had employed previously are now able to focus on the 5 % exceptions that actually need to be reviewed by a human. But now it’s become the use of artificial intelligence for exception management. We haven’t gotten rid of any people. We’ve now been able to identify more of that. And if anything, we’ve probably invested in more people if we found a higher level of exceptions, especially earlier in those deployments.

Cris Bjelajac (24:11)

Yep.

Adam Parks (24:23)

as we’re refining the models for the purpose.

Cris Bjelajac (24:26)

Yeah, wonder, I guess I question whether or not actually people are even doing 100 % of the calls. They may have tried it for a little bit, but they’re scaling back and going back to the 5%. I think that’s operationally where a lot of the, where our potential customers may be. Even though you can do 100%, depending on the, I mean, I don’t, I don’t.

Adam Parks (24:50)

From an AI perspective, you don’t think they’re listening to 100 %? Interesting.

Cris Bjelajac (24:56)

Yeah, and we’ve seen those sort of leaps over the years where technology came in and said, well, we can do a QA and we can look at those types of things. And what happens is you see, and this is less collections, but more on my old Genesys side, on the contact center side of things.

We would tout, yeah, you can do 100 % or you could double or triple or quadruple the amount of QA that they’re doing. From a statistical standpoint, it was not beneficial to do more.

but you were using tool keyword tools and things like that listening in real time to and flagging calls, flagging interactions. And so that was a different sort of a different model. I don’t see people doing a hundred percent of the calls unless they’re unless they’re, know, they’re and customers asking for them, but I haven’t seen that either. So I’m dubious that that is a use case. Yes. Is it where

money is being spent and being made by providers. Not really. I don’t think so. If I was starting an AI company, that is not where I would go to a…

Adam Parks (26:05)

and not where you would focus your time. I

respect that, makes sense. third party collection agencies also had the broadest deployment of virtual negotiators and SMS workflows. So it seemed like that was where they were investing in those communications. Now to me, when I think about the virtual negotiator and I think about the SMS workflows or email workflows, it brings me back to the general idea of self-service.

Cris Bjelajac (26:11)

Yeah.

Yes.

Adam Parks (26:31)

I think the consumers are looking for self-service opportunities. They want to self-serve. They don’t want to talk to anybody. They want to keep it simple. But the virtual negotiators and you we can get really good at sending text messages, but we better have a really good frictionless portal or the text messages don’t add significant value or the ability to convert. It’s like.

Cris Bjelajac (26:34)

Yes, absolutely.

Adam Parks (26:51)

It’s like being able to run great social media ads and you you sell me on this product on Instagram and ready to buy it. I click it, your website looks like it’s from 1990 and I’m going, not putting…

Cris Bjelajac (27:01)

Yeah, yeah, you still need to that that those are table stakes that still need those you got to consistently improve those types of things. People but you’re absolutely right. The negotiation is that’s where I see a lot of value and where our customers are seeing a lot of value is the negotiation engines are better. You know, they’re

Adam Parks (27:01)

I’m

Cris Bjelajac (27:18)

They’re faster, more customer, the customer experience is better. So, you you want to make a friction, you want your, you want to be paid first out of that bucket of money and you want to make it as frictionless as possible and you want to use AI to help enable that. And if you’re not, and I mean, that’s where everybody’s spending their time. That’s where I’m having conversations with our customers. They’re like, hey, I’ve just demoed a bunch of these different platforms.

this is the one I want to integrate to. And I’m like, okay, well, tell me why. Because we’re looking at them, we’re all kind of impressed with them, but we’re making different value judgments on who we want to spend time investing in. And they’re like, well, this one is doing this and it gives me this operational efficiency of X, Y, and Z. And I’m like, okay, all right, that’s great. I would rather spend time and money integrating what my customers want than what I think my customers need. So when somebody comes in to me,

with real data that’s saying this one is the right one, Cris, it makes sense for me and then I therefore I think it makes sense for you guys too.

Adam Parks (28:20)

Here’s the challenge with the right one. That’s where I really struggle right now because 12 months ago, ChatGPT was the right one. And then Perplexity got better and that became the right one around June. And I started going, my God, I’ve been using the wrong thing for six months. And then Gemini pops out of nowhere and breaks every benchmark across all of them on pretty much every front. And so that constant changing, I feel like the AI race has just begun.

Cris Bjelajac (28:23)

Yeah.

Adam Parks (28:46)

And it’s hard for us to predict, I think looking at it, the way that you’re looking at it, I think is the best way that we can look at it. But I think it’s a constant evaluation process of what’s the best one today.

Cris Bjelajac (29:00)

Well, I’m looking at the, I look at these small vendors that are the startups that are coming to all of our customers and our friends that we see that we hang out with at the shows and stuff. And these are all, you and I were up in Vermont a few months back and there was a vendor there that just decided just hadn’t.

Adam Parks (29:11)

Yeah, sure.

Cris Bjelajac (29:23)

been at a show before, was thinking that collection was where she wanted to go. She was just, you know, was shaking hands and making moves and things like that. But she, know, you still need to…

meet the criteria of this industry as a vendor, which means you can’t be fly by night. You have to come in and invest. You have to show that you have staying power. You have to show that you can walk. If you don’t have experience in collections, but you’re willing to learn and, you have people on staff who’ve been in collections for a those are all still important, regardless of the technology. And you can’t come in and carpet bag this business. You’ve got to spend and invest and be there for the

for the years. getting back to your point, the underlying big engines and how these small startups use those engines, are they building their own engine? I don’t know if that’s a good use of money.

Adam Parks (30:25)

So

that comes down to the size of the organization, right? Like I talked to EXL, Mike Walsh and their team are, they have their own models, they have their own stuff from the ground up, but it’s a three or $4 billion company. And then for a startup to come in and try and build a model from the ground up. And I think it’s also about the commitment, because you brought up a really good point that there’s a lot of exploratory conversations, people coming in making grandiose statements.

Cris Bjelajac (30:35)

Yeah.

Adam Parks (30:50)

At ACA this year in Louisville, I met a group who was gung-ho. We’re going to be the voice company for debt collection. I was two months later at the DCS conference. They’re like, well, we’re not so sure that we want to do voice. We might want to build a system of record or we might want to do this or we might want to do that.

Cris Bjelajac (30:50)

Yeah.

So, and that was our advice six to nine months ago, is let’s see how they shake out. We don’t wanna spend a lot of limited time on integrating one of these new startups to latitude unless they show some staying power, some effective sales. So we’re judging these vendors, not so much on their technology and their capabilities, but there’s…

What does their balance sheet look like? Who’s invested behind them? What is their time? Are you going to be here in six months? Because as we just spoke a little bit ago, a lot of those guys didn’t, they didn’t last. didn’t. Now I get it. The pressure on those, these new vendors is immense. They need to show wins to their investors as quickly as possible. we have to

Adam Parks (31:37)

the money in.

Cris Bjelajac (32:02)

We have to figure out those value judgements. when a new vendor comes to us and says, hey, I’d like to integrate, by the way, there’s no way we can handle all the companies that want to integrate to latitude. It’s just too many. But we’re not just evaluating technology. We’re evaluating those other things, the business side of things. Are you going to be here? Because we have to make smart bets along those lines.

Adam Parks (32:28)

I think you guys are looking at it the right way. No, I think this is exactly on topic because when we start talking about how the different companies are using it, I think that’s kind of part of the conversation is what does that criteria look like and why are they making some of these decisions? Because when we talk about law firms, for example, law firms are prioritizing compliance, documentation, integrity, and scalable review.

Cris Bjelajac (32:32)

Yeah.

Adam Parks (32:53)

So they’re using it for operational efficiency. They’re trying to leverage it from a documentation doc prep standpoint. Like I think they’re taking some of the things that they’ve used BPO’s for in trying to apply some technology and having the BPO staff manage the exceptions from the process, from an AI perspective. So it feels like that’s a different animal.

Cris Bjelajac (33:12)

Great use, great use, yeah.

I agree, and I like those types of, so to speak, rudimentary. You’re using AI to do stuff that you used to have a person do, and you can retask that person and do much higher value stuff.

Adam Parks (33:24)

Fundamentals.

Cris Bjelajac (33:30)

you know, lay that off on an AI engine and let it go. If it’s not 100 % of what you used to get now and it’s only at 80%, you’re still way ahead of the game. And by the way, it’ll probably be at 100 % three months from now. I mean, we’ve talked about the improvements. you’re listening to this and you haven’t played around with

you know, at the various engines like the ChatGPT, the difference between ChatGPT six months ago and the, and the 5.1 releases is night and day. it was frustrating as hell to use the old version of ChatGPT. it. Yeah. And I, and by the way, I don’t blame anybody who was trying it back out maybe even two years ago and going, is this, I don’t know, but it is.

Adam Parks (34:09)

3.0 was very frustrating.

Cris Bjelajac (34:21)

remarkably different experience and I expect it to be a remarkably different experience six months from now, know version X version Y version Z coming out. yeah, I ⁓

Adam Parks (34:31)

big part of that is the context

window. How many tokens can it understand within the context window? And I think that’s where the larger the context window, the more valuable it becomes for larger projects. Like 3.0, could use it to rephrase an email. Like you could use it for some basic things. But as the context window expands, the capability of understanding a particular task and accomplishing that task at a high level, I think increases.

Cris Bjelajac (34:34)

Yes.

Adam Parks (34:59)

Now, when it comes to law firms, what I thought was interesting is they’re the most cautious adopters of AI. So they’re kind of the slowest on the curve of adopting. And they also seem to be the least happy with the overall investment. Meaning if we ask them, does your AI investment exceed, meet your expectations, is it falling a little bit behind? I the law firms are the least likely to be optimistic about it. Maybe that’s just the legal profession, or maybe that is, you

because of the use cases that they’re specifically deploying within their organizations or the IT resources that they have to support this type of operation.

Cris Bjelajac (35:32)

You know, it’s interesting you say that about the law firms because the, I’ll talk a little bit about my former employer, Genesys. think where one of the business units within

Genesys that saw a lot of operational benefit was in the legal side of things, contracts and things like that. So my question to the law firms might be, you may not be experiencing on sort of the front end on the customer or I mean, the debtor side of things, but are using it operationally for other things or if you’re not, I would be hard.

Adam Parks (36:02)

think that’s the case. They’re doing less in terms of consumer communication, more in terms

of back office, more in terms of documentation management, but overall they’re slower to adopt than other organizations. Because when you compare it to creditors, creditors were slower to adopt in the first couple of years in which I had the survey data, but now they’re building in-house.

Cris Bjelajac (36:09)

Yeah. Okay. Yeah.

Yeah.

Yes.

Adam Parks (36:24)

Banks are not looking at it from that perspective. I mean, we had the webinar recently where we were talking with Arvest Bank and how they’re making their technology driven decisions, which I think is a great window for anybody who is in debt collection and is interested in understanding how the banks are viewing technology. That might be my favorite conversation of 2025 to date.

Cris Bjelajac (36:34)

Yes.

It

was so different from, you know, the talking to banks back in 2005. was such a, God, it so depressing trying to say, okay, SMS kit, you know, people love it. want to, they want to get reminders. it’s like, I don’t know. And it just, know, year after year, just trying to, you know, flying down to, you know, these, these areas and demoing it and every, yeah, it’s just like,

Adam Parks (37:10)

Yeah, well physically having to go there.

Cris Bjelajac (37:15)

Come on guys, but it’s totally different now, I think. I mean, I see the law firm, yeah, they’re lawyers, they know, they understand the risks and they’ve been through it, right? If they’ve been in the business for a while, they’ve been through it. So I don’t blame them for being more cautious, I just keep, I’m gonna keep, you have other subjects who don’t feel this way. I just don’t see it. I just do not see the…

fear that we used to see. I just don’t see it. Yeah.

Adam Parks (37:42)

I think it’s lessened, but ⁓ from

a creditor standpoint, I don’t think it’s a fear. think they’re just getting hyper-focused on what I would call enterprise approved third party AI solutions. The creditors are looking for something that has already been proven in the marketplace. They know it’s going to work and their focal points in terms of deployments, like in terms of use cases has been QA, really customer experience. So chat and written communication. And I think they’re exploring voice very hard.

but also that omni-channel orchestration because they’ve got so much data going back to the origination of the account. They understand the consumer better than anybody in third party collections because we don’t see all of that data. I don’t think all of that intelligence is being transferred to us when an account goes charge off. So the banks and the originating creditors, I think just have a better visibility into

how this consumer is gonna act, which gives them more capability from an omni-channel orchestration perspective.

Cris Bjelajac (38:42)

Yeah, at the end of day, are they doing anything fundamentally different on the various, on the collection? When you get right down to it, collecting of debt, you you send a Dunning letter, you do this, the steps, the workflows that you take, are you doing anything fundamentally different if you are an early AI adopter or a heavy AI adopter?

No, you’re not really doing anything funnily. You’re using it to tune it and improve it and you’re having operational, you know, you’re retasking people. Maybe you’re not replacing some headcount on a natural attrition. You’re replicating your best agents across this, you know, new use of technology. But are you doing anything super fundamentally different that would raise regulatory flags or anything like that? No.

Adam, before I, there was something that’s being asked of me lately that I wanted to ask you and I wanted to know if you’re hearing about it and then maybe anybody who’s listening. One of the things that is being asked of me is from a technological standpoint is redundancy.

within using these AI tools like an agent assist. Excuse me, not redundancy, used the wrong word, latency. Does this new AI assist have the speed, the compute to…

provide a good customer experience or is it get, you I think we’ve all been sort of like having an, doing an international call where you sort of talk over each other and the latency becomes a problem. Yes, exactly. But we, you know, we ran into that back in the day in the old dialer.

Adam Parks (40:14)

like we’re doing an international call right now.

Cris Bjelajac (40:22)

like did you could you do the automated voice could it listen to the text to speech and then and then play the right audio prompt this is very similar here yeah

Adam Parks (40:29)

So interesting, interesting question.

actually wrote an article about this not that long ago. You can find it on my LinkedIn profile where I specifically talk about AI latency from this perspective. Here’s what I’ve learned recently in conversations with these groups and in testing is that there was a time where it was and it still is, but I think we’re getting to the cusp now, especially as we go to these larger context windows. There was a time where it was speech to text to the model.

back to speech, I’m sorry, back to text and then back to speech. So you had all of these steps in the process, but AI is getting to the point to where it’s speech, model, speech, and we don’t necessarily have to go through that transfer process. So if we’re not already there, we’re within six months of speech to speech going into the model. And I think that changes everything. I don’t believe that right now in the high-end systems, the enterprise level systems, that there’s enough latency to be noticed.

I think things are happening so fast. Now on the flip side of that, if you’re using artificial intelligence for like language translation, I think it’s a little bit different. And I’ll use the example I’ve got Apple AirPod version three, I’ve got the latest Google pixel buds and I’ve got the meta glasses. I’m down here in Brazil. I don’t speak fluent Portuguese and none of these are useful for me. Portuguese happens too fast. I have every latest and greatest tool available. None of them are worth the wait.

Cris Bjelajac (41:26)

I agree.

You’ve tried, you’ve tried, you’ve tried to have conversations. Is it late?

Adam Parks (41:55)

They’re all paperweights. I try them every day. I try to understand my mother-in-law.

It’s like I got nothing. No, the language goes too fast and it can’t pick up on all of it. And so the translations that I get back are garbled ridiculousness and they’re not saying the same thing. Right? Like I was trying to, we were having a conversation at dinner the other night. So it’s a little bit latent.

Cris Bjelajac (42:03)

Hahaha.

And is the latency also bad? Well, the translation is bad,

but is the latency bad? Yeah.

Adam Parks (42:22)

A little bit, I mean, again, literally the latest and greatest Samsung Galaxy Fold 7. is the best chip on the marketplace for this type of technology at best, even if I use it directly from the device.

Cris Bjelajac (42:31)

Yeah.

Okay, so you’ve got, so the question is, that latency caused by, or is that a, is that latency at speed? Okay, so.

Adam Parks (42:40)

I think it’s a problem with the speech. think Portuguese happens so fast

that the microphones can’t pick it up and can’t understand it fast enough to respond to it. And it does take time for the latency depending on the internet connection because it’s not doing all of that translation natively on any given device. Even though they’re all touting it, right? Like Apple, when they released the AirPods, the Google Pixel Buds, it’s on every ad I get on Instagram, but none of them are meeting the expectation of actual use.

Cris Bjelajac (42:47)

Okay.

Yeah.

Got it. All right.

So let’s talk about the next use case. So let’s say you have a contact center platform. You can go as high as 200 concurrent conversations because you have 200, let’s say you have 200 agents. And.

you’re instructing your agents to use the agent assist. are not, they’re not actually, AI is not doing the conversation. And you are communicating via APIs to your vendor who is listening in real time and providing the guidance back to the agent. The latency between that, I mean, I don’t think, I think what you said before was accurate, the big platform.

Adam Parks (43:46)

That seems to be smoother.

Cris Bjelajac (43:49)

Yeah, the big platforms, if you’re using the native app within the big platforms and they’re hosted on a hyperscaler and they have the compute capability, it’s not so much of an issue. getting back to the small vendors, they don’t have the money to buy the compute power to make this happen. So you can get a demo of this and they might have…

crack the code on a better customer experience, but if they don’t have the compute power to run it and you’re introducing latency, that’s where I wonder if you’re hearing anything out there about that, because that’s what’s being asked of me.

Adam Parks (44:27)

think you are spot on. And honestly, I think you’re spot on. I think it comes down to the level of compute power. What kind of investment, if you look at the cost of Nvidia chips and all of the hardware equipment that’s necessary to support these types of things, if they’re not built on a pre-existing model, they’re either borrowing the compute power from an AWS, from a ChatGPT, or whatever their flavor is. But I think that’s where we have that disconnect right now. And Cris, I think this is a perfect

This has been a perfect conversation. I mean, we’ve covered so much information as it relates to all of the AI use cases that are specific to debt collection, how they’re being used. For those that want to learn more and see the stats behind some of the convert that drove our conversation today, the TransUnion Debt Collection Industry Report for 2025 will be released at the RMAI Annual Conference in February.

Cris Bjelajac (44:58)

We’re done?

Adam Parks (45:19)

Right before the keynote, TransUnion is gonna release it. I highly suggest, even if you’re not at the conference, download the report, read through some of the statistics because we do break it down based on company type. We break it down based on company size and how those investments are having an impact, how organizations feel about their AI investments and just so much information as it relates to artificial intelligence. But Cris, thank you so much for coming on today, sharing your insights with me. I always enjoy our conversations.

Cris Bjelajac (45:20)

that report.

Sam here, thank you very much for having me and I’m looking forward to that TransUnion report. I always love watching the changes over time on those graphs, it’s great.

Adam Parks (45:59)

This

year you’re gonna see even more in terms, I got more data this year because this is my second and going into my third year of conducting the report. So I think we’re gonna get better and better at kind of some of this trending information. But for those of you that are watching, if you have additional questions you’d like to ask Cris or myself, you can leave those in the comments here on LinkedIn and YouTube and we’ll be responding to those. Or if you have additional topics you’d like to see us discuss, you can leave those in the comments below as well. And I bet you I can get Cris back here at least one more time to help me continue to create great content for a great industry.

Cris Bjelajac (46:25)

Ha!

Adam Parks (46:28)

But until next time, Cris, thank you so much for your insights. really appreciate it. And thank you everybody for watching. appreciate your time and attention. We’ll see you all again soon. Bye.

Cris Bjelajac (46:33)

Thank you guys.