How lenders are using tech to show a more human side

If you received a text from your cred­it card provider recent­ly, you may have noticed that the tone of it seemed a lit­tle friend­lier than that of pre­vi­ous mes­sages.

The way that finan­cial insti­tu­tions are com­mu­ni­cat­ing with cus­tomers has changed, dri­ven in part by the Finan­cial Con­duct Author­i­ty (FCA), which has urged them to do more to sup­port the many UK con­sumers whom it deems vul­ner­a­ble. Research by the FCA has clas­si­fied 27.7 mil­lion adults as hav­ing some “char­ac­ter­is­tics of vul­ner­a­bil­i­ty, includ­ing poor health, expe­ri­ence of neg­a­tive life events, low finan­cial resilience and low capa­bil­i­ty”.

The issue is about more than mak­ing text mes­sages to cus­tomers sound more cor­dial, of course. Com­pa­nies are using sophis­ti­cat­ed tech to gath­er and analyse data about cus­tomers, so that their staff can offer tai­lored advice and sup­port. Debt-col­lec­tion com­pa­ny Lantern uses tools sup­plied by Max­Con­tact, for instance. These can draw infor­ma­tion about any giv­en customer’s his­to­ry from numer­ous sys­tems and present it to the firm’s call-cen­tre agents while they’re talk­ing to that indi­vid­ual. 

“All this data gives us a pic­ture of the cus­tomer. It answers ques­tions such as: have they default­ed before? How often? Do we have details of their income and expen­di­ture? From there, we can offer a bet­ter ser­vice,” says Lantern’s CEO, Denise Cross­ley. “For instance, there would be no point in get­ting some­one to agree to some­thing they couldn’t afford, because they’d sim­ply end up in default. This is about estab­lish­ing the sort of rela­tion­ship in which cus­tomers will call if some­thing goes wrong because they know we’re here to help them and pause pay­ments if we need to.” 

In some cas­es, firms can inte­grate their own infor­ma­tion on a cus­tomer with third-par­ty data for fur­ther insights into an individual’s poten­tial vul­ner­a­bil­i­ty. For instance, REaD Group offers a sys­tem, which is used wide­ly in the util­i­ties indus­try, that bases its mea­sures of vul­ner­a­bil­i­ty on aver­age data in a par­tic­u­lar post­code cov­er­ing fac­tors such as age, income, socioe­co­nom­ic sta­tus, employ­ment and long-term ill­ness rates. 

Going one step fur­ther, some lenders are using a form of AI called nat­ur­al lan­guage pro­cess­ing to pre­dict a customer’s vul­ner­a­bil­i­ty based on their calls, emails and webchats. This soft­ware tracks words and phras­es along with audio cues such as the pitch, tone and pace of a con­ver­sa­tion. It can then crunch thou­sands of pieces of data to alert users at the first sign that a cus­tomer is start­ing to become dis­tressed and/or con­fused. 

At Lantern, the Max­Con­tact sys­tem scans con­ver­sa­tions for par­tic­u­lar ‘lan­guage trig­gers’ from cus­tomers that can prompt an agent to react in a spe­cif­ic way or share cer­tain infor­ma­tion. Cross­ley cites an extreme exam­ple: “If some­one were to men­tion sui­cide, that call might be flagged to a senior man­ag­er, who could offer sup­port.”  

This approach helps to reas­sure and mol­li­fy cus­tomers, but it also makes busi­ness sense for lenders, she adds. “In our mar­ket, what hap­pens in the rela­tion­ship between the com­pa­ny and the cus­tomer ulti­mate­ly affects the pric­ing of our debt pur­chas­es. If we pro­vide a bet­ter cus­tomer expe­ri­ence, we see low­er default rates and high­er recov­ery rates. This in turn enables us to price more com­pet­i­tive­ly in the future.” 

The Key Group is a provider of equi­ty-release mort­gages to peo­ple aged over 50. Any­one sign­ing up for such a loan is required to meet one of the company’s approved inde­pen­dent advis­ers to check that the prod­uct is suit­able for them. The com­pa­ny recent­ly invest­ed in a speech analy­sis tool from Aveni, which scans audio record­ings of these meet­ings to iden­ti­fy any poten­tial vul­ner­a­bil­i­ty issues. 

“These meet­ings can last up to 90 min­utes, so it would be a huge task for some­one to lis­ten to all of those record­ings,” notes Key’s CEO, Will Hale. 

The Aveni sys­tem con­sid­ers fac­tors such as the lan­guage, pace and clar­i­ty of con­ver­sa­tions. It then gen­er­ates a report that will iden­ti­fy any­thing of poten­tial con­cern. While age isn’t con­sid­ered a vul­ner­a­bil­i­ty fac­tor, the soft­ware does look for signs that a cus­tomer may not under­stand what’s being explained. If that hap­pens, the advis­er can arrange to speak to the cus­tomer again with a rel­a­tive present, or refer them to one of Key’s so-called vul­ner­a­bil­i­ty cham­pi­ons – agents with advanced-lev­el train­ing in how to deal with cus­tomers who may need extra sup­port. 

Over the next six months, Key will be rolling out a new cus­tomer rela­tion­ship man­age­ment (CRM) plat­form in its con­tact cen­tre. Then it will use Aveni on calls to pro­vide instant sup­port to its front-line staff. 

“Ulti­mate­ly, we want to get to a point where we have a full account his­to­ry inte­grat­ed with our CRM and tele­pho­ny sys­tems, so that we have full AI sup­port of all com­mu­ni­ca­tions,” Hale says. “It’s very much a tool that sup­ports agents and it helps to ensure that we aren’t approach­ing con­ver­sa­tions with the wrong pre­con­cep­tions.” 

Although the sys­tem­at­ic detec­tion of vul­ner­a­ble cus­tomers can be help­ful to those indi­vid­u­als if it means that they receive more sup­port, peo­ple are enti­tled to pri­va­cy, stress­es Simon Thomp­son, head of data sci­ence at IT con­sul­tan­cy GFT. 

Finan­cial insti­tu­tions that use this type of analy­sis in con­tact cen­tres need to estab­lish prop­er pro­ce­dures spec­i­fy­ing how the results will be used, he argues. “You can’t just stir this data into the mix and hope for reli­ably pos­i­tive results. You need to mon­i­tor and review both the inter­ests of the cus­tomers and how well this tech­nol­o­gy serves them.” 

It’s vital that peo­ple are aware that their con­ver­sa­tions are being analysed and the extent to which the analy­sis affects the firm’s deci­sion-mak­ing, Thomp­son adds. “Ulti­mate­ly, tech­nol­o­gy can tell you only so much. Does a cus­tomer sound dis­tressed on the phone because of the way their query is being treat­ed, or because a tod­dler is jump­ing on their feet? We all need to remem­ber that there’s no such thing as an eth­i­cal machine. The ethics always rest with the peo­ple and the organ­i­sa­tion. It’s a respon­si­bil­i­ty that can­not be out­sourced.”