Members who are struggling don’t necessarily want your human touch

Most credit unions don’t realize their members would rather talk to an AI agent about falling behind on payments than to a human.

The shame, the judgment, the fear of explaining what went wrong, all the things members want to avoid when they don’t call can be fixed. Every time credit unions lose the opportunity to help because the member never picks up the phone.

Catherine York Powers, CEO of Constant AI, is building a different kind of loan servicing automation. Constant AI helps credit unions make the member experience about solving problems, not denying requests.

Her company’s flagship product, Nia, is an AI agent that handles payment skips, deferrals and due date changes. A member calls and says, “I can’t make my payment.” Nia asks a few questions, checks the credit union’s rules, makes the decision and writes it back to the core in minutes, not days. No applications. No waiting until Monday while the member gets more delinquent.

Nia is designed to mimic a conversation with a human loan servicing agent. If a member isn’t eligible for a skip, Nia pivots. Can they make a partial payment? Are they eligible for deferment? Can we solve this another way?

The automation handles 80% to 90% of cases. The hard conversations, the ones where credit union staff actually need to understand the situation and apply judgment, those go to humans. The technology frees up staff to solve tougher problems instead of processing routine requests.

For examiners and compliance teams, every step is auditable. The transcript, decision logic and consent are all documented and exam ready.

The future isn’t apps; it’s conversations. It’s saying, “Skip my car loan payment” and having it done by an AI bot, not clicking through screens. Credit unions that embrace this shift first are going to win with members who’ve been too embarrassed to ask for help.

Listen to Catherine’s full conversation with Sarah Snell Cooke on this episode of The Credit Union Connection podcast to hear how automation works, why governance is a top priority in AI and what the future of member service will look like.

Sarah Snell Cooke: Hello and welcome everyone. I am Sarah Snell Cooke, your host here at the Credit Union Connection. I am joined today by Kathryn Powers. Welcome.

Catherine York Powers: Hello, Sarah. Nice to see you.

Sarah Snell Cooke: Great seeing you as well. And Kathryn is the CEO of a company called Constant AI. Why don’t you tell us a little bit more about yourself and the company?

Catherine York Powers: Yeah, absolutely. I am the founder and CEO of Constant AI, and our focus is to democratize loan servicing automation for all credit unions, small and large. Lately we’ve been doing that by enabling this kind of automation. So think about payment skips, deferrals, due date changes, all of those sorts of things that are super complex, require precise rules, read and write to the core, all of those complex bits.

And we’ve been able to push that functionality up through AI chat, AI voice, either our own in the form of Nia, who is our agent, or through our partner API, so that if a credit union already has an agent, say with Eltropy or Posh or others, then it can be consumed that way as well.

Sarah Snell Cooke: Mm-hmm. Tell us a little bit more about Nia.

Catherine York Powers: Yeah. So Nia is the first ever AI skip-a-pay agent. If a member calls their credit union and Nia is sitting behind that institution, a member can simply say, “Hey, I can’t make my loan payment,” or, “I wanna skip my loan payment.” Nia will ask them a couple of questions, go in and grab the rules for the credit union, make sure that member is either approved or denied based upon those rules, and then write everything back to the core.

So there’s no waiting, no applications, no forms to send in, those sorts of things. And we think that’s really important because if you request a payment skip, say, on a Thursday, but nobody gets around to it till Monday, you could actually fall delinquent, and therefore not be eligible for that, and then roll into delinquency, and nobody wants that, right? Not good for the member, not good for the credit union.

And I think more importantly, I’m actually behind on the website, but we have upskilled Nia already, so she’s not just doing payment skips, rather she’s mimicking a conversation a member would have with the loan servicing agent, just saying, “I can’t make my loan payment.”

So if they’re not eligible for a payment skip because they’ve already gone 30 days past due, for example, Nia can say, “Well, hey, can you make a partial payment?” And if not, she’ll check to see whether they’re eligible for a deferment or extension, which is usually a hardship accommodation, and then write that back to the core as well.

So it’s not like you can’t get a payment skip conversation and you hit a denial wall. Rather, our goal is to solve the problem.

Sarah Snell Cooke: Mm-hmm. Yeah, that makes perfect sense. It’s not… it is just rule following, but at the same time, it’s a little more human. So speaking of which, is there a way, like for whatever reason, that a human would want or need to break in to the conversation?

Catherine York Powers: Yeah, absolutely. First of all, if a member actually wants to talk to a human, they can certainly say, “Please transfer me.” We totally respect that, as do all contact centers or service companies. But there may be something more complex, right? The delinquency is too severe where the kind of support the credit union would provide is not yet automated, or they wanna learn more about the situation before they offer the right solution.

So our goal with the automation underneath is just to take those workflows that can be automated, can read and write back and forth to the core, and handle that. So let’s say 80 to 90% of the cases, but the whole goal behind the use of this technology is to free up credit union staff so they can solve those tougher issues and spend time with members.

Sarah Snell Cooke: Mm-hmm. And speaking, I served on the board of my credit union, and I know we switched to a text solution that would text people when they’re behind on their loans, and people seem to be more comfortable with that.

I think it’s probably less judgment, and so I think automating that can really help the member experience too because even if they aren’t being judged, they probably feel like they are. Thoughts on that?

Catherine York Powers: Yeah, I think that’s a real human experience. Privacy is the product, right? We’re all human beings.

If you are experiencing a hardship, the last thing you wanna do is talk to a stranger about that and then be asked questions about, “How did this happen,” and, “What are your spending habits?” and all of these sorts of things. So I think the privacy and humanity behind—strange to say, but humanity behind an AI agent, whether it’s served up in chat, if you’re getting a text saying you’re 10 days past due, having that redirect to a chat interface or a voice interface to support that fuller conversation and help resolve the challenge is, I think, an incredible step forward, especially for loan servicing teams that have been handling this manually in the past.

Sarah Snell Cooke: And you’re working with MSUFCU. They’re always known as being ahead of the game. How’s that going?

Catherine York Powers: Yeah, they’re fantastic partners, and Reseda Group is also an investor in Constant. So they’re early adopters of Nia, and also they are great with spending time with companies like ours to help us think about broader applications and roadmap and what are the things we should be thinking about to expand the use cases.

So they’ve been incredible partners. On serving a specific institution, they’ve been great there. We also, on the partner side, have partnered with Eltropy. They’ve been amazing partners too, and I think helping to prove out this thesis that credit unions that have already deployed their own AI voice or chat agents and having us sit underneath to power that is an approach that makes sense for those credit unions that already have relationships with these established players.

So if you think about this, we’ve separated these two jobs where I think many AI systems mash together. The language model’s only job is to understand what the member is saying. Never actually touches the decision. The actual decision—can this member skip a payment, get a deferral, change their due date—for us, runs in this separate deterministic engine.

So basically, the rule book that always gives the same answer for the same situation. So no improvising, no guessing kind of thing. And we think that infrastructure separation is really important from a security standpoint as well. So think of it as like pilot and autopilot. The AI pilot is talking to the passenger, calm, conversational, answering questions, but it’s not the one flying the plane.

It’s the autopilot, our decision underneath. So whether it’s powering Nia or powering these other voice agents offered by these other companies, it is an infrastructure that is secure from a governance perspective, protecting member information. But more importantly, it democratizes loan servicing automation for anyone.

So if you’re a $100 million institution using Eltropy, for example, we can power that through Eltropy. Or if you’re a $13 billion institution, it doesn’t really matter to us as long as we’re sitting underneath a company like Eltropy to deliver that.

Sarah Snell Cooke: Mm-hmm. And you mentioned governance. That’s super important right now, and I think, talk a little bit about the key points to governance that credit unions should keep in mind when they’re discussing AI.

Catherine York Powers: Yeah, I think a lot of the focus is on the front door, right? How do we make sure there’s not synthetic fraud? How do we make sure the member is who they say they are?

And there are companies like Illuma that are doing a really good job with that, partnering with the Eltropy, Glia, Posh of the world to make sure that front door is secure. And our focus is, once they get through that front door and they’re in the house, how do we make sure they only get into the rooms they’re allowed to get into?

And so that separation, we call it a state machine, but that separation underneath where all of the rules are read exactly like the credit union wants them to read. So same input should generate the same output is how we secure it underneath, and separating those layers from a governance standpoint I think is important.

And we decided to attack this from the infrastructure standpoint. There are others that are doing a great job doing this—like Eltropy has a safe AI, right? And so there’s transparency with everything, so you can—not speaking about Eltropy specifically, but all of those companies are making sure that you’ve got visibility into the conversation, into the transcript, making sure that you have all the layers of security that are necessary, securing the APIs, all of that.

I think governance is really important. I saw a report recently that there are more and more chief AI executives being appointed because governance is such a huge thing, right? So deploying these agents is increasing the attack surface, so you’ve gotta make sure that when you choose a provider, that you’re spending time on making sure governance is in place so that whatever actions are being taken are defensible to examiners, are defensible to your internal compliance team.

Sarah Snell Cooke: That was gonna be my next question about auditability. So when an examiner comes in and you’ve decided to decline a skip-a-pay or move the member from a skip-a-pay to deferment, things like that, how does the examiner know you went through, or Nia went through all the right steps?

Catherine York Powers: Yeah. So I’ll talk about this just from Nia’s perspective versus partners. A few things. One, the transcript is available. The audio recording is available, so you know that person actually said yes or no to the terms and conditions, for example. But underneath, because we’ve separated those two technologies, everything related to the rules is locked down in our state machine.

So just like we’re doing today, powering a million and a half transactions in digital banking, we’re able to show the actual—we call them applications internally—with every step that was taken along the way, where the member consented, the loan ID, the reason for decline, all that information is available to the compliance teams and examiners.

And actually, we are beginning to build an agent that is going to create exam-ready packages so that credit unions don’t have to go and look around and say, “I have all this information now, I’ve gotta put it in my own package so it’s exam ready.” We’re working on that. We’re excited about that possibility to help with governance and making credit unions feel more comfortable with these decisions.

Sarah Snell Cooke: Mm-hmm. And I assume that then that’s also not just member-facing, but the reports themselves are agency-facing, regulator-facing as well, which you know that they know what they’re looking for, so you gotta have it together. But yeah, that’s super important to save time for credit unions, absolutely, which is one of the big complaints, especially from the smaller ones.

You mentioned earlier, speaking of size, credit unions from $100 million to $13 billion, is there an ideal size for a credit union as far as where they are gonna have the maximum ROI on this type of tool?

Catherine York Powers: Yeah, I think from a financial standpoint, obviously I think the larger institutions with larger loan portfolios that are looking to grow without hiring additional staff or hiring too many additional staff, this makes a lot of sense for.

I think the smaller institutions struggle because a lot of times they only have one or two people that are servicing all the loans, servicing all of the collections activity. So there’s more than ROI to be considered for those folks, right? And so it might be a little bit more expensive, but when they think about the opportunity cost of not doing it, there’s more to consider than whether or not the dollars make sense.

And also, those are the ones that are looking to grow and scale, and it’s hard to scale a loan portfolio without hiring more people. It’s always been the way things have been done to service or collect on those loans. So I think the impact may be felt more in a smaller credit union, but in terms of real dollars ROI, I think as you scale up, this makes a lot of sense as well.

Sarah Snell Cooke: Yeah, and that’s equally important to a smaller credit union too, is to save on other expenses. And so I always allow guests to leave with their final thoughts. What would you like to leave our credit union audience with today?

Catherine York Powers: Yeah. I am so excited to watch this movement from a GUI to LUI, right? The GUI, meaning the app screens you’re clicking through, to LUI, the language user interface, where the app itself becomes secondary and the conversation is the interface. So I’m really excited to watch what’s happening in digital banking because I think the future is—and I’m not the only one saying this—instead of opening the app and finding the transfer button and picking the account and hitting confirm, or finding the skip-a-pay button and clicking through screens, being able to say, “Hey, move $200 to savings,” or, “Skip my auto loan payment,” I think it means that the app is going to disappear behind these intelligent conversations, unless you’re looking to just see your balance, right?

No one’s gonna wanna type, “What’s my balance?” It’s right there on the front screen. But for anything else, I think this is where the change is gonna happen, and it’s going to make what we’re doing, I think, even more important, because having these agents do stuff, not just answer questions, is where the future is, and where the impact is really felt both for the member and the credit union.

Sarah Snell Cooke: Yeah, love it. Thank you so much for your time today, Catherine. Appreciate it.

Catherine York Powers: My pleasure. So lovely to see you.