Using AI to cope in the coronavirus era

1. Sales prioritisation

Sales and busi­ness devel­op­ment are suf­fer­ing and AI-pow­ered sales per­for­mance solu­tions can help. So-called propen­si­ty mod­els can iden­ti­fy which cus­tomers are most like­ly to buy a prod­uct or ser­vice from a com­pa­ny, says Dr Tom Dav­en­port, pres­i­den­t’s dis­tin­guished pro­fes­sor of infor­ma­tion tech­nol­o­gy and man­age­ment at Bab­son Col­lege, Mass­a­chu­setts. These mod­els can help those work­ing in sales improve their pro­duc­tiv­i­ty and effec­tive­ness, by show­ing them which cus­tomers to pri­ori­tise.

“For brands, hav­ing insight into what their cus­tomers think and want has always been a key pri­or­i­ty, but the COVID-19 pan­dem­ic has made this under­stand­ing even more crit­i­cal,” says Chris Col­ley, prin­ci­pal of cus­tomer expe­ri­ence at Medal­lia. But he notes, at the same time, col­lect­ing data on what cus­tomers think has become more chal­leng­ing.

As peo­ple stay at home, con­sumers have shift­ed from per­son­al inter­ac­tions, where they pro­vide direct feed­back, to dig­i­tal inter­ac­tions. “Instead of vis­it­ing a bank branch where they can speak to the cashier, they are more like­ly to be bank­ing online,” says Col­ley. “As it’s no longer pos­si­ble to eat out, they’ll be order­ing online deliv­er­ies. The same pat­tern is being repli­cat­ed in sec­tors across the board. This shift is cre­at­ing a ton of new, unstruc­tured data, which can be hard to make sense of.” That’s where AI solu­tions can cut through the noise and find out what con­sumers feel and need.

2. Matching demand and supply

Com­pa­nies are inter­est­ed in match­ing demand and sup­ply, and that’s going to be real­ly crit­i­cal com­ing out of this cri­sis,” says Dav­en­port at Bab­son Col­lege. “The good news is there’s more and more exter­nal data avail­able on demand.” A big steel com­pa­ny, for exam­ple, has infor­ma­tion about the var­i­ous fac­tors that might influ­ence demand for steel, such as the demand for auto­mo­biles. These demand mea­sures depend on exter­nal data that’s used to match up to what their sup­ply chains can pro­duce. “So that you’re not pro­duc­ing more than you need to sat­is­fy demand and you’re not leav­ing unful­filled demand out there,” he says.

AI solu­tions can analyse this exter­nal data. But, as Dav­en­port points out, AI typ­i­cal­ly relies on data from the past, while the COVID-19 cri­sis is unprece­dent­ed. There­fore, com­pa­nies have to ensure they use data that is rep­re­sen­ta­tive. He says: “I sus­pect that in some indus­tries, the past will be a bet­ter guide to present and future activ­i­ty than it is in oth­ers.”

Dav­en­port notes that we do have data from the 2008 finan­cial crash to go on, but the cur­rent cri­sis is hap­pen­ing at a faster rate and its con­se­quences might dif­fer. There are, how­ev­er, some indus­tries where spend­ing pat­terns might con­tin­ue at sim­i­lar rates, such as gro­ceries and con­sumer sta­ples. “Peo­ple have to buy deter­gent, no mat­ter what,” he says. Con­verse­ly, expen­sive con­sumer goods might strug­gle. Data from the last finan­cial cri­sis might give some indi­ca­tion of how much demand there might be.

3. Document and identity verification

AI can work on iden­ti­ty and doc­u­ment ver­i­fi­ca­tion, says Dr Ter­ence Tse, asso­ciate pro­fes­sor of finance at ESCP Busi­ness School. Think of a bank, for instance, that needs to ver­i­fy its cus­tomers for onboard­ing and com­pli­ance. This is often done by human check­ers, who check payslips or dri­ving licences. “It’s a very cost­ly, inef­fi­cient process,” says Tse.

Instead, AI can be used to “quick­ly iden­ti­fy the type of ID doc­u­ment cap­tured, deter­mine if the secu­ri­ty fea­tures of the ID are present, per­form face-match­ing – com­par­ing the pic­ture in the ID to the per­son in the self­ie – and even help deter­mine whether the per­son is phys­i­cal­ly present”, says Robert Prigge, chief exec­u­tive at Jumio.

“For the past few years, dig­i­tal account open­ing has been at the top of the list of tech­nolo­gies organ­i­sa­tions intend to add or replace, but COVID-19 is push­ing this ele­ment of dig­i­tal trans­for­ma­tion to the front of the line,” says Prigge.

4. Back-office tasks

AI-pow­ered cog­ni­tive assis­tants can per­form a company’s back-office tasks. This includes order­ing new cred­it cards, issu­ing refunds or can­celling orders, says Faisal Abbasi, UK man­ag­ing direc­tor at Ipsoft. He notes: “When the cog­ni­tive assis­tant is unable to han­dle a task due to its com­plex­i­ty, this can be seam­less­ly hand­ed over to human agents to man­age. This ensures the time of those team mem­bers is spent solv­ing the most chal­leng­ing prob­lems and focused on val­ue-add activ­i­ties.”

This process is often referred to as robot­ic process automa­tion (RPA) and is increas­ing­ly com­bined with machine-learn­ing. It spans all sorts of back-office ser­vice oper­a­tions, as long as they are struc­tured tasks, such as automat­ing the claims process­es of insur­ance com­pa­nies or banks.

“Almost all the com­pa­nies that I talked to about RPA said, ‘Oh, we’re just using it to free up peo­ple to do more cre­ative, less struc­tured work’,” says Dav­en­port at Bab­son Col­lege. But he notes that if the cur­rent COVID-19 cri­sis leads to a severe reces­sion, which seems like­ly, com­pa­nies will use it to replace work­ers. “My guess is that it’s going to con­tribute to sub­stan­tial job loss­es or at least slow­er growth of employ­ment after the reces­sion because com­pa­nies will have auto­mat­ed a fair amount of work,” he says.

5. Cash-flow forecasting

Over the next few months, cash flow is like­ly to con­tin­ue to be a seri­ous con­cern for small­er busi­ness­es as rev­enue streams dry up. But there are a num­ber of fore­cast­ing AI solu­tions that can help. “Cash flow is always an issue in dif­fi­cult economies,” says Bab­son College’s Dav­en­port. AI solu­tions are already in place that analyse data for the pur­pose of cash-flow fore­cast­ing.

One impor­tant caveat is “you have to make sure you have the right data peri­od to cre­ate mod­els that would be use­ful for this cur­rent envi­ron­ment”, he says. Once again, AI can only help if the data we feed it is rep­re­sen­ta­tive.

“You have to go back to reces­sion­ary envi­ron­ments to ask, what were your cash needs in the past? And again, it’s dif­fi­cult because this reces­sion appears to be hap­pen­ing much faster,” says Dav­en­port. Eco­nom­ic data comes in slow­ly and a reces­sion is typ­i­cal­ly defined as two quar­ters of neg­a­tive GDP growth. He adds: “We won’t have this data until the end of June. But I think there is not much doubt among econ­o­mists that we’re in a reces­sion already.”

6. Medical support

The COVID-19 cri­sis has put unprece­dent­ed pres­sure on NHS staff as pub­lic health has tak­en cen­tre stage. “Med­ical ser­vices have been ter­ri­bly shak­en and our beloved NHS may be near a coup de grâce,” says Dr Alex Ribeiro-Cas­tro, data sci­en­tist and senior teach­ing fel­low at Impe­r­i­al Col­lege Busi­ness School in Lon­don.

He says health tech may offer a tem­po­rary buffer to allow non-crit­i­cal ail­ments to be treat­ed, leav­ing clin­ics and hos­pi­tals free to focus on crit­i­cal cas­es. An exam­ple is Doc­tor­link, which pro­vides online doctor’s appoint­ments and has algo­rithms that can pro­vide med­ical­ly endorsed diag­nos­tics. Anoth­er is Baby­lon Health, which is build­ing an AI-based health app that can help diag­nose patients’ issues. It’s effec­tive­ly a chat­bot that can “trans­late layman’s lan­guage into med­ical ter­mi­nol­o­gy and deduce what may be caus­ing the pain”, says Ribeiro-Cas­tro.

Dinesh Venu­gopal, pres­i­dent at Mpha­sis Direct & Dig­i­tal, says: “AI-based chat­bots and robot-advi­so­ry ser­vices can very well be use­ful in reliev­ing the admin­is­tra­tive bur­den on extreme­ly busy and under-resourced health­care staff, automat­ing process­es such as screen­ing patients for symp­toms and record­ing nec­es­sary infor­ma­tion.” By reduc­ing the amount of face-to-face inter­ac­tion between patients and hos­pi­tal staff, this goes a long way to less­en­ing the risk of spread­ing infec­tion, he says.

7. Staff demand, supply and infrastructure

Giv­en that many employ­ees may have to self-iso­late dur­ing the COVID-19 out­break, AI can analyse the num­ber of staff need­ed. “AI com­pa­nies get requests from

their clients to iden­ti­fy if they are like­ly to even have enough work­ers to staff a rail­road,” says Dav­en­port at Bab­son Col­lege. In this case, AI can help to match demand and sup­ply, but from a labour stand­point. “If com­pa­nies are lay­ing off peo­ple, they’d like to know it’s the right num­ber of peo­ple. Mak­ing sure you have enough peo­ple to staff a par­tic­u­lar train or a pro­duc­tion shaft could be quite dif­fi­cult.”

Trans­porta­tion com­pa­nies rep­re­sent a sig­nif­i­cant com­po­nent of a country’s infra­struc­ture. “They are faced with an unfor­tu­nate Catch-22 sit­u­a­tion: we, as a soci­ety, need to keep crit­i­cal infra­struc­ture and its employ­ees healthy, how­ev­er not all of them can man­age crit­i­cal infra­struc­ture remote­ly,” says Ribeiro-Cas­tro at Impe­r­i­al Col­lege Busi­ness School.

What’s more, semi-automa­tion is already imple­ment­ed in cer­tain forms of pub­lic trans­port. Ribeiro-Cas­tro cites Navya, a com­pa­ny that designs and man­u­fac­tures autonomous vehi­cles, such as shut­tle bus­es at air­ports or theme parks. “AI is already being used more gen­er­al­ly in the trans­porta­tion sec­tor to do things such as increase pas­sen­ger safe­ty, reduce traf­fic con­ges­tion and acci­dents, lessen car­bon emis­sions, and also min­imise over­all finan­cial expense.”