Editor’s note: This interview has been edited for length and clarity.
Sarah Wolak: Coming out of the MBA supply chain event last month, there was a lot of discussion about supply chain pain points that could be solved with tech. What has ICE seen about specific borrower pain points it aims to address with the new agents?
Matt Dowd: There are two different people we deal with: the borrower and the user. But skills and performance are at odds with each other to some extent.
We’ll start with the borrower experience. There are usually well-defined reasons why a borrower engages with their employer. They want to understand their payment, maybe why their payment hasn’t cleared, or why it has changed. Escrow is great. At tax time, it says “where are my tax documents?”
Those are usually quick answers. But today, you have to pick up the phone, call someone and possibly sit on hold. That is not a good experience for the borrower.
Through chat or voice, borrowers can ask those questions, take action on their credit and get answers 24/7, instantly. If they need more explanation, they can go deeper. So whether it’s a conversation or a voice, we have those precautions, and we know enough if someone asks a question that should be passed on to someone, it will.
So that’s very important to the borrower’s experience, because it allows them to ask the questions they want and get an answer 24/7. And then when they need someone, they can ask those initial questions, be directed, and all that paperwork is filled out by a customer service agent.
SW: So everything is written down to the agents?
MD: Yes, all of our AI is controlled, and what really sets us apart is that our AI technology is built into the system of record. Think how important that is: It’s not something you buy as a bolt-on or bundled-together service. Whether you’re using an MSP for customer service or automation or service, digital AI is built into automation.
SW: What was the previous option for borrowers who wanted to contact service providers? Was it a call center model?
MD: Yes, of course. When we talk to our customers, they have very well-defined metrics, common reasons for borrowers to call. And most of them are pretty basic – like, real people calling and saying, ‘How come my payment can’t be processed?’ or ‘When is my next payment due?’
The big thing is that people’s fees go up because of escrow, and people always wonder why it went up. Instead of having this question at 9 o’clock at night and the borrower doesn’t want to wait until tomorrow, they can ask the AI, and it will give them a clear, concise answer.
SW: And sometimes, as a customer, you don’t really want to take the time to call with one question.
MD: Yes. One thing we all know is that buyers, the longer they hold on, the more frustrated they become. That’s agnostic in everything, and it’s very difficult to work in a call center because you don’t really know when you think your busiest times are. You don’t know how long that person will wait and you can’t predict their state of mind [the customer is in] before they enter.
It’s all about allowing the borrower to engage with employees on their own time, on their own channel. Some people want to just talk to a person, though, so our technology is built so that there’s always what I call an exit route, which means they can be directed to an agent as soon as they want.
SW: I wanted to ask about the concept of controlled AI since ICE deals with a lot of sensitive client and server information. How is that managed, and how are agents trained to keep that information as safe as possible?
MD: It’s basically the same way we keep our data secure today. One thing about ICE is that we focus on data privacy. But our clients, they can do whatever they want with their data. If they wanted to take it out and use it elsewhere, as they do today, they could still do that. But our models, are always dominated. Everything we do is readable and understandable. But again, we won’t provide public LLMs with all this proprietary data.
SW: Going back to the rough spots, what about the service? How do these tools help them succeed?
MD: By providing information on AI tools, the employee gets the same thing: They don’t have to take as many calls. That said, as those conversations come in, they are monitored and tracked. So if a customer calls a month later, we’ll use our AI to give a prediction as to why they might call.
On the voice side, the employee benefits, not only because they reduce their costs by taking fewer calls, but also because it is better for the borrower since they have a 24/7 service to help them.
SW: Finally, to what extent does this affect the relationship between the borrower and his employee, as these tools aim to further establish or maintain that relationship?
MD: The borrower’s relationship with the servicer is often thin during normal times, especially for autopay borrowers who rarely interact with their servicer. But back to your question about the relationship: If the borrower is a major homeowner and is paying automatically, the relationship is probably good.
But the relationship changes when the pay rises. That relationship becomes more critical when borrowers start having financial problems. When someone gets into trouble or starts to get into trouble, that borrower’s relationship with the servicer becomes critical. Now they are in danger of losing their home, and it is the user’s responsibility to do everything possible to keep them there. So how well an employee performs against that will make or break their relationship with the client.
But since borrowers don’t have a choice in who works, it’s an interesting relationship because there’s less control for the borrower. It is even more interesting if it is a company that provides services but also exits. If you service and originate, that relationship with the borrower is important if they’re going to refinance, do a HELOC or buy a second home with you – if you maintain a relationship and get maintenance.
