Can emotion AI really boost sales?

The sales rep might not be the only one watch­ing you on your next vir­tu­al meet­ing. Some com­pa­nies are employ­ing ‘emo­tion AI’ – a sub­set of AI that is able to detect human emo­tions – to mon­i­tor indi­vid­u­als on sales calls, pro­vide feed­back on their reac­tions and high­light the most engag­ing parts of the pitch. 

The mar­ket for such sales enable­ment plat­forms is grow­ing, with con­sul­tan­cy firm Ver­i­fied Mar­ket Research claim­ing it could be worth $7.3bn glob­al­ly by 2028. 

Zoom is one of the lat­est entrants to this bur­geon­ing mar­ket, launch­ing Zoom IQ for Sales last month. Described as a con­ver­sa­tion­al intel­li­gence soft­ware, it claims to offer “mean­ing­ful and action­able insights from cus­tomer inter­ac­tions to improve sell­er per­for­mance and enhance cus­tomer expe­ri­ences”.  

How­ev­er, the announce­ment was met with oppo­si­tion from a num­ber of human rights groups and pri­va­cy cam­paign­ers. An open let­ter to Zoom CEO Eric Yuan – co-signed by the Amer­i­can Civ­il Lib­er­ties Union and the Elec­tron­ic Pri­va­cy Infor­ma­tion Cen­ter – called for the video com­mu­ni­ca­tions com­pa­ny to halt its plans to advance the fea­ture. “This move to mine users for emo­tion­al data points based on the false idea that AI can track and analyse human emo­tions is a vio­la­tion of pri­va­cy and human rights.” Zoom declined to be inter­viewed for the piece. 

Accord­ing to Uniphore, one of the com­pa­nies behind the tech­nol­o­gy, its emo­tion AI assis­tant ‘Q for Sales’ can “help sell­ers ‘read the room’, sense emo­tion­al cues, and improve engage­ment”. 

Patrick Ehlen is vice-pres­i­dent of AI at the com­pa­ny. He thinks the fears over emo­tion AI come from a mis­un­der­stand­ing of the tech­nol­o­gy.  

“Per­son­al­ly I’m not a huge fan of the term emo­tion AI,” he says. “There seems to be a con­cern that when peo­ple talk about emo­tion AI, they’re talk­ing about a com­put­er sys­tem that can actu­al­ly read your inter­nal emo­tion­al state with a high degree of accu­ra­cy. The truth of the mat­ter is there’s no AI sys­tem that can do that.” 

 Getting AI to understand context

Psy­cho­log­i­cal research backs up this asser­tion. The way that peo­ple com­mu­ni­cate emo­tion is not uni­ver­sal and facial expres­sions are much more con­tex­tu­al­ly sen­si­tive than pre­vi­ous­ly thought, accord­ing to a 2019 paper from psy­chol­o­gy pro­fes­sors at the Cal­i­for­nia Insti­tute of Tech­nol­o­gy, North­east­ern Uni­ver­si­ty, Uni­ver­si­ty of Glas­gow and Uni­ver­si­ty of Wisconsin–Madison. 

Lisa Bar­rett, pro­fes­sor of psy­chol­o­gy at North­east­ern Uni­ver­si­ty and one of the paper’s authors, believes the find­ings “call into ques­tion the sci­en­tif­ic jus­ti­fi­ca­tion used by some emo­tion AI tech­nolo­gies”. 

Ehlen is there­fore reluc­tant to make any bold claims about the technology’s capa­bil­i­ties. He main­tains that “it’s an unre­li­able tech­nol­o­gy to be mak­ing very big, crit­i­cal deci­sions”. How­ev­er, he thinks it could prove use­ful on sales calls. In sit­u­a­tions where a sales­per­son may be speak­ing to mul­ti­ple peo­ple on a video it can be hard to deter­mine who is engag­ing and who isn’t.  

“You’re look­ing at peo­ple in this tiny lit­tle win­dow, so it’s not as easy to read the room and see what peo­ple’s facial expres­sions are,” he says. “If you can have a machine gaug­ing their reac­tions and deter­min­ing at what point the CFO seems inter­est­ed, it can be a help­ful tool.” 

To do this, the visu­al AI soft­ware analy­ses a num­ber of ele­ments of the con­ver­sa­tion, includ­ing facial expres­sions, tone of voice and people’s ges­tures. “From that infor­ma­tion, we’re able to get much clos­er to hav­ing a 360-degree view of what peo­ple are doing when they’re in con­ver­sa­tion to bet­ter under­stand them,” Ehlen adds. 

The plat­form can then give sales­peo­ple real-time feed­back on sen­ti­ment and engage­ment to help them adapt their respons­es and, in the­o­ry, improve their sales con­ver­sions. 

Behavioural AI 

Sybill is anoth­er plat­form that claims to use emo­tion­al intel­li­gence to accel­er­ate the sales process. How­ev­er, co-founder and CEO Gor­ish Aggar­w­al is reluc­tant to say the pro­gramme can iden­ti­fy people’s emo­tion­al state. “We con­sid­er it as a behav­iour AI, which is dif­fer­ent from emo­tion because emo­tions are very sub­jec­tive,” he says. “You can­not tell whether a per­son is fear­ful, angry or con­temp­tu­ous just by look­ing at their face.” 

The soft­ware instead looks to iden­ti­fy an indi­vid­u­al’s body lan­guage or facial expres­sion to high­light key moments, Aggar­w­al says: for exam­ple, whether some­one is nod­ding along to a con­ver­sa­tion or smil­ing. Sales­peo­ple can then revis­it the record­ing to see when peo­ple were most engaged and which parts they should fol­low up on. 

Although Aggar­w­al claims its AI can deter­mine whether some­one is nod­ding their head with up to 95% pre­ci­sion, the technology’s impact in the sales con­text is yet to be proven. Both Uniphore and Sybill’s plat­forms launched ear­li­er this year and are both cur­rent­ly con­duct­ing val­u­a­tion stud­ies to deter­mine to what extent their AI pro­grammes can improve sales per­for­mance. 

Jason Bell is asso­ciate pro­fes­sor of mar­ket­ing at Saïd Busi­ness School and works on AI mod­els for com­put­er vision and nat­ur­al lan­guage pro­cess­ing. He believes there is a lot of ‘over­claim’ in the emo­tion AI mar­ket.  

“Many web­cams and front-fac­ing smart­phone cam­eras are low qual­i­ty, so even if you have a great pre­dic­tive mod­el, the sig­nal that it’s get­ting is not great,” he says. “In prin­ci­ple it’s pos­si­ble, but I haven’t been con­vinced by the cur­rent tech­nol­o­gy.” 

Rely­ing on this tech­nol­o­gy in a sales con­text could lead peo­ple down dead ends, accord­ing to Bell. His biggest con­cern cen­tres on the technology’s accu­ra­cy. “It dra­mat­i­cal­ly over­sim­pli­fies emo­tion­al states,” he adds. “If you’re mon­i­tor­ing one or both par­ties involved in the sales process and cat­e­goris­ing their emo­tions, it can be quite crude … You could cre­ate more com­pli­ca­tions for the sales process with­out adding a tonne of val­ue.” 

Con­cerns around pri­va­cy, as raised in the open let­ter to Zoom, have also made peo­ple wary about imple­ment­ing the tech­nol­o­gy. Uniphore and Sybill are both aware of the con­cern. 

“We’re very aware that some peo­ple see AI as being some­what creepy,” Ehlen says. “So, we’ve tried to build in as many pro­tec­tions and safe­guards as pos­si­ble to make peo­ple feel com­fort­able.” 

As a result, Q for Sales is an opt-in expe­ri­ence. With Sybill, users are noti­fied that “the call is being record­ed and notes are being tak­en,” although there is no ref­er­ence to AI unless the user decides to reveal this.  

The iterative process of using AI

Aggar­w­al doesn’t rec­om­mend using the pro­gramme for inter­nal calls or in sit­u­a­tions where there’s a pow­er imbal­ance. For exam­ple, using it as part of an inter­view process would be “eth­i­cal­ly not cor­rect as they wouldn’t be able to refuse”. 

There’s also a risk that peo­ple change their behav­iour when they’re made aware of being record­ed. Ehlen admits this could mean “the AI is not going to work as well as it did before” and will have to go through a process of retrain­ing.  

“It’s an iter­a­tive process,” he says. “As these tech­nolo­gies make their way into the main­stream peo­ple will become more com­fort­able with them and this will become less of an issue.” 

Whether these con­ver­sa­tion­al AI pro­grammes become a stan­dard part of the sales process remains to be seen, but con­vinc­ing cus­tomers to be mon­i­tored by AI might be the hard­est sell.