How Agentic AI is Transforming Collections with Earlier Outreach and Higher Recovery

Mike Lawson here with CE Broadcast. Welcome back to our show. And on today’s program, you might recognize this fine gentleman, mister Sri Nj from Interface dot a I. Is it interface AI or interface dot a I? What’s the official name there, Sri?

Dot a I.

Interface dot a I. Which is the URL as well? So easy enough to get to. Yeah.

You guys got two birds with one stone there. So but great to see you as always, Shri. How are things going at Interface dot a I these days? How are you guys doing?

Good. Good. Well, since we’ve spoken to you, come a long way, it’s been an exciting journey. I mean, with everything going on in AI, it couldn’t be at the better place at a better time. Right? No doubt. You know, it’s exciting times for sure.

The you know, the exciting part is that I fundamentally believe with all the things happening in AI, how powerful it’s getting, we can truly solve most of the industry problems, right, using AI.

You know, we are going to talk about one of them today.

I think so. Well, thank you for doing the introduction there, Sri. So so today, you guys have built the industry’s first agentic AI collections solution called Smart Collections, and we’ll bring up a little graphic for everyone to check out here while we talk about it. So Smart Collections, as you guys can see there. So so just to kick on kick us off here, Sri, just give us the elevator speech on what is Smart Collections? What have you guys what have you guys built here?

Yeah. So so smart collection is a fully AI collection agent that has ability to follow all the rules and do outbound call, text, and email where, you know, including very soon, it can even send physical mailers to your members or or part of the early delinquents, right, which is zero to thirty days, and ensure that the recovery is at least two to three x better than a human doing a follow-up. Right? So that’s what it does. It’s a multichannel agentic AI assistant that follows up, ensures the payments are made on time.

Yeah. Do you wanna talk about the the agents real quick? These these three specialized agents that kinda work in concert to reach the members in their preferred channel at the at the right time, the right place, all that good stuff. You wanna talk about those guys real quick?

Yeah. Yeah. Absolutely. So I think collection largely is a extremely regulated space where you had to follow a lot of rules.

Right? You know, you have to follow the number of times you contacted someone, the day of the time, the disclosure you have to do at the beginning of the call. There is a lot of rules you need to follow to make sure you’re not liable. But who’s good at following rules?

Not human beings. We are known to be creative beings.

But AI is amazing following rules. Right? So, you know, what it does is it has ability to send out a quick let’s say, you know, some sometimes we all been there. We miss out a payment. We didn’t plan the cash flow properly.

Imagine, you know, Mike, you are you or I, that’s the way we get a quick text message saying, hey. Seems like you missed out a a payment here within a couple of days after your payment was due. You appreciate that. Right?

So, you know, you’re you’re, you know, you’re able to click on that message and quickly make a payment and move on with your life. And sometimes, you you know, you prefer the AI calling you to say, hey, you know what? It’s like a little bit of an escalation beyond the text if we if it if the AI doesn’t hear from you, it kind of tries to call you because not everyone is very well versed with text messages. You know, they still prefer calls.

So the AI then escalates a little bit, calls you, say, hey, Mike. Here’s your payment due and things like that. Right? So typically, happens in a a credit union or a community bank is, you the staff doesn’t follow-up with you until thirty days after that the payment was due.

Right? We you know, but, research has shown us, you know, many times and our our own data kind of shares shares this with us is, if you’re able to follow-up within the first fifteen days, your probability of recovery is two to three x higher. So but typically, what happens in a credit union is they don’t have the bandwidth and staff, availability to do the follow-up and they wait for thirty days. Right?

Yep.

Yep. And guess what? Every day passes, after the payment due, the probability is dramatically decreasing. Right?

So and especially, it’s pretty low after thirty days. Right? So with AI, it is able to immediately follow-up right after the payment due date and do a multichannel. It can even send you an email, text message, call, including shipping a physical mailer to your home saying, hey.

This is a reminder. You have a payment due and things like that. It does all of that. It orchestrates all of this outreach to ensure, you know, there is you’re not breaking any any rules.

Right? So Yep. You as a financial institution are not liable. And at the same time, it uses a safe space for a member to discuss, explore options.

Because when sometimes you’re talking to another human being about debt and you’re unable to make a payment, you know, you know, it’s it’s not an easy conversation. There’s a lot of judgment. You know, you worry about the kind of question to ask. Right?

So having an AI is a safe space that you can ask any question. No question and dumb question. Right? You you can really go through the real challenges and find options and things like that.

So it’s actually huge for the members too. Right? So and it’s a it’s a big problem in industry, Mike.

We as a credit union is about two trillion dollar industry. Right? And we have now reached the highest degraded asset quality in the last twelve years ever in the industry. Wow.

Right? So there is about, you know, twenty billion dollars out of the two trillion are delinquent loans at at this given point in time, And fifteen billion is usually what charged off out of that, you know, twenty billion, which is Yes. Almost close to seventy eight percent. Right?

So it’s charged off as a a loan that you can’t ever recover. That’s a lot of money putting back the numbers and and everything. Right? So just an effective AI solution could easily help recover easily four to five billion of that fifteen billion before the charge off.

Right? So Yep. That’s the magnitude of the problem.

Yeah. I’m gonna bring that graphic back up here real quick. Because it the yeah. Because there’s the three you can see that at the bottom of the page there, the three channels that you guys use to get to the member where they are with. But the ninety five basis you wanna talk about the ninety five basis points real quickly? Because that’s that seems like one of the the whys you guys created this Yeah. This this this solution.

Yeah. Let’s take an example. Right? Let’s say you’re a two billion dollar credit union with one point five billion loan portfolio at about one percent delinquency rate, which is right now is about one point zero two percent delinquency rate what what we’ve seen according to NCUA.

That’s about fifteen million dollars, of your portfolio is delinquent. Right? And typically, right now, the charge off is about, point eight percent. Right? So which means you’re writing off close to eleven to twelve million out of that fifteen million that, you know, you could never recover. That’s a lot of money.

Oh, yeah.

So with AI, you’re able to, you know, improve that recording by twenty, twenty five percent, especially if you put AI to follow-up on hourly delinquency. And that results in, you know, you know, awarding close to two to three million dollars of losses. Right? As soon as you you’re able to put that back into serving your members and paying the dividends or improving the products and things like that.

And it’s not about just the recovery of the loan, which is like, you know, for a two billion dollar financial institution, you can now record two to three million more. But it’s also the cost of doing that. Right? The the personal cost and the liability that you’re exposed to if the person makes mistakes and things like that.

Right? So so that’s how the math works out. So a a two billion dollar financial institution could with AI could record two to three million dollars for the delinquency rate of close to one percent.

Wow. That’s pretty cool. So I wanna get into how it works, and we’re gonna do a quick demo for for everybody out there. I’m gonna bring in your colleague, John Manning. John, welcome to the show, sir. How are you?

Great. Thanks, Michael.

You are you are welcome. So I’m gonna bring up the gonna share your screen here and show us the demo. There it is right there. So, John, if you wanna walk us through it and and share with us how this this works exactly, I think everybody will be very appreciative.

Yeah. Let let’s just do a quick overview of of what customers could take advantage of. So where I’m at right now in the dashboard view is really your collections command center. So you’re able to see everything that’s going on. So whether you’re logging in in the morning or looking at at the end of the day, you can see how much is past due and what your statistics are and how you’re getting through the the pipeline of those different accounts that you’re working with. Okay. Down here, you can see kind of the activity of what’s been going on, and you can actually get into even the the narrow granular specifics of that activity of who’s it’s who it’s been reaching out to, the different channels it’s reaching out to, and just getting a better understanding of what’s going on throughout the day.

Through from here, what we can do is actually go and and look at some of our accounts and what’s going on. And I’m just gonna pull up an account here that we can see is past due on a credit card. We can see all the amounts and the specifics and the details. The really key critical things here are getting into what is the preferences of that particular member.

You know, what are they allowing? What have they defined in those conversations? If they mention, hey. Don’t call me again.

The AI can recognize that and update their preferences. So it’s really intelligent when it’s having those conversations, and it it’s not just limited to having to log in to a web page and and check those those preferences off. That’s cool.

Yeah. Also within here, you’re gonna be able to see all the interactions that AI has actually had with this member. So here you can see it verbatim. What what was sent to them via SMS or email?

And so if anyone ever calls to complain and say, you know, hey. I already paid this. Why do you keep calling me? You can go and you can actually look at the logs of what’s taken place and if those conversations have even gone out.

That’s really slick. From here Track and I’m I’m I apologize. So this is all trackable. Right?

So Oh, a hundred percent. The reporting and analytics all behind it actually give you great visibility into understanding, you know, who’s paying, who’s not paying, getting a better sense for what channels are actually producing better results. And so that way, you can you can help the AI target those specific areas and Sure. Actually become more effective and more efficient.

Yep. Yep. Very cool.

And then, of course, you can manually override a lot of these things. So that’s that obviously there for for control purposes. But, really, this is meant to be autonomous and being able to allow it to go out and do the collection on your behalf. And this can be designed for all the different types of products you may offer your members.

And, you know, you may decide to use it for some versus others or all of them. It’s very flexible in how you do that. Yeah. And you can just see the examples here.

And each of those products can have a different type of journey on how you’re doing that collection. So if we were to just look up an example journey that we have here, it starts with a simple SMS. Back to Sri’s example of, oh, maybe I forgot to make my credit card payment. Shoot.

A soft f SMS reminder.

I’ll go in and make that payment, and, you know, that was a benefit to me because it just was something I forgot to do. I was busy that month. So being gentle in the touch and then escalating that to a voice conversation, sending emails, It can escalate through any type of journey that you design. And you may already have some best practices that you’re using today. You can then use those best practices to help build those journeys, and you can assign it to the different types of products and how you wanna do that collection based on what you’re you’re collecting on.

Very nice. Very because obviously when you guys know this all too well and every credit you know knows that’s all too well that, you know, dealing with collection is a very delicate situation.

And I think the more information you have, especially ahead of time, the better off the situation is gonna be. And and, yeah, again, this goes back to, like, what you talked about earlier, Sri, is like, you know, the early stage delinquency is becoming a a sharper focus across the industry. You wanna address that really quickly here before we before we move on with the demo? Just because because this really helps you get into it early. Right?

Yeah. Pretty much. Right? Like, it’s the next day after the payment was due. Right? Like, you know, it is the AI is on it.

It’s following up, sending reminders, and, you know, it it can escalate as the day goes by. Right. But the important part is, you know, know, this is this is happening with no effort from the financial institution. Right?

Because today, they wait until thirty days because the the the the staffing challenges and, you know, in collection and the lab with the exposure, they’re like, I don’t wanna take the risk. Let’s just wait here for thirty days. If they don’t make it, I’ll call. But guess what?

Like, you know, the, you know, the the property of collection actually decreases significantly after thirty days. Right? So so yeah. You know, now you don’t have to worry about staying compliant.

You don’t have to worry about hiring people. You let the AI do, like, immediately and then day one after the payment due. They can make sure it’s effective and the recording is, like, easily two to three x higher than you would have done after thirty days.

Oh, I can imagine. Yeah. For sure. I think your members gotta be so appreciative as of as well, you know, especially getting in there early.

Oh my gosh. Before the before the real trouble starts. Yeah. So yeah. For sure. For sure.

John, you wanna I apologize for interrupting the demo, but that question just kinda popped into my head. So I had Yeah.

Great question. And and just to kinda the other aspect of this, think, is really critical and important to call out is, you know, the compliance aspect. So, you know, being able to follow the guidelines, all of that is built in. So these federal guidelines are automatically updated via API.

So, again, this is designed to be autonomous. You’re not having to oversee it and monitor it. It it is designed to go out and help collect some of that collect some of that for you. Yep.

And then in addition to that, you may have your own independent policies of how you’re collecting. And so you can go in and define those on your own, and that will stack on top of those federal guidelines. And so you have full autonomy and control over what the AI is doing, and you you’re building in those guardrails around it.

And and every financial institution may have a little bit of a little a different philosophy in the culture and how they reach out to their members, how often, and things like that. Maybe a little different from the federal policy. Right? So you can go in and say, you know what?

We wanna take a softer approach with our members. Here are the phrases that are prohibited. Right? So we can call that out clearly and the AI is personalized.

There is and it’s a tool reflection of the financial institutions, you know, values and, you know, philosophies. Right? So so it’s not like one size fits all. I think you can personalize to make sure it is following the culture and the values that the financial institution care about.

Very nice. Very nice.

Yeah. And to that personalizations, there’s also the ability to design what are those those messages that are going out, whether it’s an SMS, a phone call, or even an email. Or or to Sri’s point, some of the actual mailers that we’re gonna be introducing later, having the ability to go in and tailor those based on how you wanna send those out and how the that wants you want that to be received by your membership base.

The last thing I’ll touch on here before we actually go into a real example is just what you touched on earlier about the analytics and reporting. As you can see, there’s a lot of robust visuals and data that you can collect from this and understand how it’s performing, how it’s doing against the metrics that you’re looking at, and just getting a sense of, you know, what channels are performing versus the other channels. So a lot of good data to be able to derive from this that you can, of course, export and pull into a BI system or maybe a different reporting tool that you’re leveraging. So lots of flexibility from the platform.

Yeah. No doubt. That’s really cool. Really cool.

So the last thing I wanna touch on before we go any further is, you know, if I wanted to manually come in here and and kinda override some of these, like I was mentioning before I have the ability to trigger, you know, what what is the daily run plan for today, and it’s gonna tell me what it’s gonna be working on. And then I can go ahead and say, well, let’s execute on that. And in this case, it’s gonna go out and make a call, and it’s actually just calling me right now. So let’s go ahead and listen to what it has has to say.

Hello. This is collections agent calling from First Community Credit Union. May I speak with John Manning?

Yeah. This is John Manning.

Thank you, mister Manning. Before we proceed, I need to verify your identity for security purposes.

Can you please provide either the last four digits of your Social Security number or Yeah.

It’s one zero eight one.

Thank you for verifying that, mister Manning. This call is an attempt to collect a debt, and any information obtained will be used for that purpose.

I’m reaching out regarding your credit card account ending in demo. It shows a past due balance of one hundred twenty five thousand dollars, and it’s now twenty seven days overdue.

How can we assist you in resolving this today?

What are my options?

You have a few options, mister Manning. You can make a payment for the past due amount of one hundred twenty five thousand dollars to bring your account current. If you’re ready to proceed, I can send you a secure payment link via SMS or email. Alternatively, if you’re facing financial hardship, we can connect you with our assistance team to explore available option.

Can you tell me more about hardship options?

Certainly, mister Manning. If you’re experiencing financial hardship, our assistance team can explore options such as setting up a payment plan or possibly modifying the terms of your account to more manageable term. I can note your request and have someone from the team reach out to you to discuss possible solution.

Would you like me to arrange that?

Yeah. That sounds great. Thank you.

And we, of course, can take that conversation and escalate it to a human agent at that time. Ultimately, the the whole goal is if it does get a connect, it gets a live answer, being able to identify what’s going on, how can we help that member. If it’s hardship, get it routed to the right skill or queue or department, and have somebody there that’s gonna be able to help them with a hardship program or help them work through that challenge that they’re dealing with.

You know, from the AI perspective, if I wanted to, I could get a payment link sent to my SMS or email and actually just autonomously take care of all this myself Very cool.

Which leads into where the member portal resides. And so being able to have that link go in and and actually make payments and manage my account profile.

Goodness. Wow. What was your outstanding balance? Was it a hundred and eighty five thousand dollar payment or something like that?

Yeah. It was a little bit too much.

I You’re gonna have made on your Lamborghini payment or something?

What’s going on there?

You bought a you bought a home in that condo. Wow.

So I I think the the important part is, right, like, you know, when you look at that call and the quality of the conversation, right, like, this is all the disclaimers you need there.

And, you know, it’s just a pretty pretty great kind of experience. You can, you know, John was able comfortably asking what are the hardship problems, you know. But if it was a human being, he probably would have little hesitation, you know, where would the other person would judge him for asking that question. That’s actually a bigger problem for the members.

They don’t wanna be in a judgmental environment when they’re talking about their financial hardship completely. So Yeah. Yeah. That’s the kind of a big part of it.

Yep. For sure. All of this ties back to our one AI brain, Mike. You know, we’ve been doing this agentic platform that can take inbound calls.

We have done numerous recordings with you here where our customers, hundreds of financial institutions are we have been able to handle incoming calls to the contact center and the AI was able to pick up the call and answer questions. And with this smart collection, it’s an outbound call. Right? This is the same engine.

It’s the same AI. We’ve, you know, process millions of interaction for hundreds of financial institution. Instead of waiting for someone to call and make a payment, now it starts to make a call itself outbound or a text message or an email.

And so this one’s half of the same AI brain that’s proven for over a decade.

Oh, this is so so cool. And the beauty of this, as you guys well know, it’s only gonna get better.

So the more it learns, the more data it gathers. Yeah. It’s gonna it’s just well, that said, how does this get better? So, I mean, obviously, it’s pretty darn slick right now.

I mean, how do you improve? I mean, I know it’s still early in the game. You guys are probably, you know, you know, getting everything, you know, straightened out here. But, you know, how does this get better?

I mean or or do we even have or can you even answer that question?

I think it’s all about the predictive model that powers the engine. So figuring out what if, say, as an AI and what time of the day you call, how what words do you use when you are calling or chatting or sending an email? What cadence works? You know, you send the email first or SMS first or a call first.

There’s a lot of modeling that happens to figure out what is effective strategy. Right? And that model is continuous to learn, so it becomes very effective in being able to ensure the collection is happening in effect, you know, in a successful way. That’s the AI that’s learning. Every time someone pays pays their amount due or don’t pay for amount after the follow-up, it is learning something.

Right? It knows what to do next time. Right? So just imagine if you have if you have a two billion dollar credit union with probably fifteen collection agents, imagine if if they’re all learning’s is connected and and and and it’s continuously that it’s, you know, you’re benefiting off of that.

Right? Yep. The collective intelligence. Now imagine there’s a hundred financial institution with tens of thousands of collection agents as we could power that they’re all exchanging the best practice and learning from each other what worked.

Right? That’s the power of the AI. Right?

So it’s learning model.

Yeah. You just answered my question big time, Sri. So thank you for that. Because I’m I’m looking at this and it’s like, oh my gosh.

This is awesome. It’s mind blowing. But like, wow. How does this get even better? And you just answered it for sure.

So so last question for you here. And John, you can you can chime or chime in as well, but I think we all know the answer here. But what’s the big benefit here to credit unions using smart collections here? Sri, you wanna go first?

The big the big benefit message here?

Yeah. The biggest benefit is you’re giving a safe environment for our members to have an important conversation about their financial That’s probably the number one benefit. And second, you’re able to achieve that with an you know, with with efficiency. Right?

So in in a in a compliant way. Right? So you don’t have liabilities and exposure. You you don’t have to hire to do it, but at the same time, you achieve a friendly way to help your member To have those tough conversation if if they’re having financial hardships.

That’s kind of, like, the most important benefit.

And and and a really super efficient way as well. Right? Exactly. Yeah. For sure.

And and, John, did you wanna add anything to that to Yeah.

I I would just say, you know, I really relate this to just giving giving the members a choice of being able to do something complex that typically had to land on a human, doing it with AI, and to Shree’s point, doing it in a safe environment. So, you know, it’s a very sensitive topic and being able to have the choice of leveraging AI, whether it’s over an SMS or an email or even doing it over the phone. Just I I relate this back to those all the channels that we try to give our members to be able to communicate over. Now let’s just enable those channels to be more effective with AI and give them options to be able to do complex tasks.

Yep. And they certainly appreciate that. I’m sure. So yeah. Good stuff. And it just makes the credit union that much more valuable, and it makes the credit union that yeah.

Just it just it’s there’s a lot of upside here, obviously. So yeah. So good stuff here. So, Shri, where can folks find out more about smart collections?

Where do we go? Obviously, interface dot a I. I’m sure you have a whole web page dedicated to it.

Yeah. So interface dot a I and you have a solution section. If you click on it, you’ll see all of our solution in the collection too.

Good deal. Good deal. Alright, gentlemen. Well, anything else here to add to wrap up here? John, did you have any more of the demo to share with us or any any final thoughts here before I let you get going?

No. I just I I’m really excited about this solution. It’s just it’s gonna be a game changer for a lot of these organizations that just haven’t had the time to really work on a collections program and just open up doors and revenue for them.

Indeed. Yeah. Well said. Sri, I’ll let you have the final word, sir.

You know, I think AI is going to make all of us more productive and enjoy the work.

It it we’re supposed to. Right? So because we always get stuck with the manual repetitive parts of our work.

But, you know, a financial institution employees having access to such a tool, now they could focus on more creative work, more strategy, or architecting a solution, which is kind of the, like, what how human beings brains are designed for, not to do repetitive work. Right? So for me, this is a freedom to the collection agents, then they can focus on higher order work. Right?

And also, this is safety for the members to have a conversation, the hardest conversations that that part of their finances. Right? So there’s a freedom and safety. Right?

Indeed. Yeah.

Right? Because I’m hearing more and more how AI is actually making us more human.

And this is a perfect example of that right here.

Exactly.

Alright gentlemen.

Well, thank you so much for your time and your insights and the very cool demo. That was a lot of fun, and I can’t wait to get this one out there and show everybody what’s capable. And but, again, this is just this is just the beginning. This is gonna be just I can’t wait to have a a conversation like this a year from now and see where we are with it. It’s gonna be a lot of fun.

Absolutely. Excited. Thank you, Mark.

Alright, guys. Yeah. Appreciate the time. Keep up the great work there at Interface dot a I, and look forward to next chat down the road. Well done, guys.