How lenders and servicers are using AI in call centers

Img

Calling a mortgage servicer or subservicer to find out details about a loan is often time consuming for both the consumer and the company representative aiming to help. 

That is why mortgage lenders and servicers are evaluating ways to streamline and automate the experience for customers. Companies operating in the servicing space are starting to dip their toes into using artificial intelligence to solve lingering call center-related problems and shave down the amount of time each call takes. 

Companies are largely cautious in their deployment of AI today, opting for functional use cases such as having an internal chatbot help a call center agent answer customer questions, or using AI to pull tidbits of information from a recorded conversation to find areas of customer support improvement.

But the overarching goal — if the regulatory environment allows it and the technology is developed — is for AI to have full blown conversations with consumers, companies interviewed said.

Sergey Dyakin, chief information officer at Celink, says the reverse mortgage servicer has instituted a number of technology initiatives to streamline its call center, including using tools that speed up a borrower's ability to reach relevant information and those that help identify customer sentiment.

"We used analytical and machine learning tools to look at the call transcripts and discovered that up to a third of our customers who successfully authenticate on [interactive voice response] are asked to repeat this process with a live agent, which added on average 50 seconds per call, impacting both the customer experience and also the agent productivity," he said. "We identified that this is a specific pain point, and we implemented changes based on those insights. Training agents to recognize system messages and having the reports helped us reduce the need for re-authentication and actually markedly improved call times."

Celink is also experimenting with using AI to rate calls.

"This AI tool scores the majority of a customer call, and it not only identifies the sentiment, but it also looks for some key situations," Dyakin said. "So for example, if the conversation mentions fraud, the tool can escalate the call for the intervention."

Also, though not necessarily AI related, Celink has slowed down its interactive voice response to serve its customer base.

"Borrowers have different communication preferences and while many of them utilize online tools, we understand that for our demographics where the average age is above 70, they often prefer traditional methods like phone calls," said Dyakin. "We have slowed our IVR system records to 85% speed for easier comprehension and have added extra wait times after each system prompt."

Dan Binowitz, managing director at Loandepot, notes the company is currently exploring a number of AI-related options to beef up its call center experience. It is in the process of choosing a vendor.

The first implementation will likely be call monitoring and recording, which could take from three to four months to get off the ground, Binowitz predicts.

"One AI application we're looking at would take all of our calls and convert them to text," Loandepot's managing director said. "It has the ability to test and consider tone, engagement, speed of conversation and provide reporting back." 

The converted text can then be reviewed "to determine whether there are areas of opportunity we're missing… is the caller interested in a new loan, or is there something else that we might be able to assist them with?" he said.

Loandepot is also considering implementing an AI-powered chatbot system to help customers on its website. Currently the lender uses "an algorithmic-based chat system that is linked into the customer portal, it is not AI as of today," Binowitz added.

Other lenders such as Newrez have also gone the chatbot route, launching an internally facing one last quarter for its call center employees. Its goal is to help call center representatives quickly find answers to caller questions.

Cenlar, a subservicer, is also in the process of developing an AI chatbot to help its agents quickly access its knowledge base and answer customer inquiries.

"Our company has hundreds of clients with different nuances in their loans, so agents have to be subject matter experts," said Josh Reicher, chief digital officer at Cenlar. Having an AI chatbot to work alongside an agent will save everyone time, he added. 

Last year Cenlar implemented a dialer machine learning system, highlighting another unique use case of AI. It analyzes patterns, demographics, and "hundreds of other metrics" to figure out when the best time to call a delinquent customer could be.

United Wholesale Mortgage earlier this year launched an AI-powered initiative to have an "overview of all of the calls that come in." 

"We could actually start to get an inflection and tone in how people were talking and it would be analyzed," said Jason Bressler, chief technology officer at UWM. "We could see if customers or team members were angry and if they were happy and pleasant, so that we could really start to change the overall training of our customer service platform." 

Lenders interviewed see a future in which an artificial intelligence bot could hypothetically have conversations with customers via phone, though most have cautioned that there are regulatory hurdles and data privacy concerns.

Binowitz said that while he is pondering the notion of one day implementing AI that can engage with customers, there is some reticence. 

"The Consumer Financial Protection Bureau has raised the concern of AI bias and that gives me a little bit of hesitation. I want to ensure that we have everything fully controlled and so that there is no bias," Loandepot's executive said.

Reicher had similar sentiments, noting that companies using a voice AI assistant have to be "mindful of information provided back" and that the company is looking further into what AI can do if its hallucinations are fixed.

"If there is confidence we would [likely move into that space]," he said. "We're just getting started, we're not at a loss of opportunities [in how AI can be implemented]." 

Bressler said that for now there isn't an AI voice assistant product out there that truly understands the financial services space and can be used with confidence.

"What will happen very shortly is that with the advent of true virtual and voice AI assistance, we'll be able to start to have AI answer calls and have human interactive conversations and then have them transferred properly and appropriately," Bressler said. "The models are not yet in a place yet where you can offer real customer service."


More From Life Style