An AI-pow­ered vir­tual as­sis­tant could be used in a va­ri­ety of ways, in­clud­ing help­ing cus­tomers to pre­qual­ify for mort­gages, eas­ing com­pli­ance and de­tect­ing prob­lems.

National Mortgage News - - Contents - By Penny Cros­man

How AI vir­tual as­sis­tants (like Alexa) could re­shape mort­gages

Al­though ar­ti­fi­cial in­tel­li­gence is not yet ca­pa­ble of han­dling a mort­gage from start to fin­ish, sev­eral lenders are in­creas­ingly turn­ing to the tech­nol­ogy for a va­ri­ety of sce­nar­ios.

Chief among them is pair­ing AI-based chat­bots with mort­gages. Rocky Stubbs, who at Cap­i­tal One helped de­velop one of the first Alexa skills for home loans, says AI will soon be more ubiq­ui­tous.

“That first Alexa Skill was very sim­ple, just, ‘ Hey Alexa, make my mort­gage pay­ment,’” said Stubbs, who is now se­nior vice pres­i­dent and head of con­sumer di­rect and dig­i­tal mort­gage lend­ing at Flagstar Bank. “The use cases for the var­i­ous sys­tems are re­ac­tive: we ask them some­thing and they com­plete one task.”

But over time, just as Siri can rec­og­nize traf­fic pat­terns and pre­dict a slow com­mute, chat­bots will start to learn about users’ fi­nan­cial pro­files and of­fer pre­dic­tions and ad­vice around fi­nances.

In­stead of be­ing siloed, the way Siri, Cor­tana and Alexa are, the next wave of vir­tual as­sis­tants will be more like Jarvis from Iron Man, Stubbs said.

“No mat­ter where [ Tony Stark] went, it was in his suit, in the car, in the house,” Stubbs pointed out. “That as­sis­tant is go­ing to be­come much more fa­mil­iar. We’re go­ing to be able to in­ter­act with it in a lot more dif­fer­ent ways.”

This kind of tech­nol­ogy could be used to pre­qual­ify peo­ple for mort­gages.

An­other use for AI and bots in mort­gages is to help fill out the ap­pli­ca­tion.

With the use of speech recog­ni­tion and nat­u­ral lan­guage pro­cess­ing, an as­sis­tant could lis­ten to a con­ver­sa­tion be­tween the loan of­fi­cer and a po­ten­tial bor­rower and start ex­tract­ing data. That could save time and po­ten­tially re­duce er­rors.

AI could also be used to help un­der­writ­ers search data­bases con­sis­tently for com­pli­ance pur­poses.

“To­day there’s tremen­dous amounts of con­tent and the guide­lines from all the reg­u­la­tors and the in­vestors that an un­der­writer needs to have at their fin­ger­tips and they take a lot of time search­ing through those to find some­thing rel­e­vant,” Stubbs said. and IBM’s Watson, for ex­am­ple, can in­gest that kind of data and an­swer ques­tions about reg­u­la­tory guide­lines.

Fan­nie Mae is al­ready us­ing AI to help pre­dict if a loan is go­ing to go into de­fault, ac­cord­ing to Mona Kahn, di­rec­tor of se­cu­ri­ti­za­tion and ser­vic­ing tech­nol­ogy at Fan­nie Mae.

In an­other use case, Fan­nie is a con­sumer of cus­tomer com­ments or “tips” made through a bot as well as through calls and emails.

“We’re do­ing nat­u­ral lan­guage pro­cess­ing work look­ing at, how do we pre­dict if that’s re­ally a true fraud tip?” said Kahn. AI can help fig­ure out what those pat­terns are.

At TIAA Bank, David An­drews, vice pres­i­dent of home lend­ing strat­egy, is build­ing out a use case for AI that helps avoid dis­ap­point­ing bor­row­ers.

“A prob­lem we had was we set the ex­pec­ta­tions up front and then we would find some­thing out two weeks down the road be­cause it was buried in some doc­u­ment a loan of­fi­cer glossed over or a pro­ces­sor didn’t catch,” An­drews said.

When the loan fi­nally got to un­der­writ­ing, the prob­lem would be dis­cov­ered — a judg­ment, say, or a lien on credit.

The bank is us­ing Blend’s ver­sion of AI and ma­chine learn­ing to un­cover such prob­lems closer to the time of the loan ap­pli­ca­tion. Loan of­fi­cers re­view this work, how­ever, as the tech­nol­ogy is still new.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.