Rare is the company today that does not claim to be “customer-centric”. Anyone unfortunate enough to have sought assistance or redress from big business may quibble. Many interactions with customer service make you feel central only in the sense of being the prime target of corporate abuse. Such experiences grew especially maddening amid the staff shortages and supply-chain snarl-ups of the pandemic. But trouble has been brewing for some time. After rising steadily for two decades, the American Customer Satisfaction Index (ACSI), a barometer of contentment, began declining in 2018. Although it has edged up from its pandemic nadir, it has shed all of its gains since 2006.
Businesses have long known that it pays to keep customers happy. In 1976 the White House commissioned TARP, a consultancy, to study the state of complaints-handling in America. Among other things, TARP’s report concluded that businesses could profit from investing more in customer service, quantifying for different industries the value that loyal customers create through repeat purchases and referrals. In the years that followed, companies from American Express to General Motors set up contact centres with toll-free phone lines to make themselves more accessible to customers. A new genre of business books extolled the value of customer loyalty. A nascent industry of consultants peddled ways to improve customer-service operations.
“In a well-functioning market, it should be profitable to satisfy your customers,” argues Claes Fornell, architect of the ACSI. What, then, has gone wrong? Increased concentration in industries from airlines and banking to telecoms could be a factor, in so far as market power weakens the will of companies to invest in pleasing their clients. Much of the consolidation in these and other industries, however, occurred before or during the period in which customer satisfaction was improving.
Technology may be a bigger part of the problem. Besides dropped calls, the most common irritation in customer-service interactions is being stuck with a chatbot, according to research from Genesys, a maker of contact-centre software. In recent years many companies have been busily deploying automation software in their contact centres in an attempt to do away with human interactions. The results have sometimes been disappointing both for customers and for companies, says Darci Darnell of Bain, a consultancy. Tony Bates, the boss of Genesys, also notes that many customer-service operations have become hamstrung by clunky old systems and messy data. That may help explain why younger upstarts, unencumbered by legacy technology, are often able to provide much slicker service, raising expectations for everybody else.
Customers, for their part, are not helping. In this year’s National Customer Rage Survey, another gauge of sentiment, 17% of customers admitted to being “uncivil” in their interactions with businesses. Scott Broetzmann, who led the survey, thinks the problem is getting worse, and that standards of acceptable behaviour are slipping. Mr Bates laments the lack of empathy customers often show towards contact-centre agents. Many such workers endure foul language and verbal abuse, one reason why attrition rates are high and rising. Staff turnover in contact centres in America hit a record 38% last year, according to SQM Group, another software provider. Higher attrition means less experienced contact-centre agents, further worsening service standards.
Could ChatGPT-like “generative” artificial intelligence (AI) make things less awful? These tools offer more humanlike interactions than earlier generations of customer-service bots. Once trained on past call transcripts and other company data, they also make fewer mistakes than the off-the-shelf version of ChatGPT, says Ms Darnell. They do a good job, too, at speedily sifting through information such as a customer’s prior interactions with a company. Whether they are an adequate substitute for humans, however, is less clear. As Jo Causon of the Institute of Customer Service, a professional body, notes, many customers are happy using self-service solutions for simple transactions but want assistance from a real person when they encounter problems.
An alternative would be to use generative AI as a complement to human agents, rather than a substitute. A working paper from earlier this year, by Erik Brynjolfsson of Stanford University and co-authors, studied the effect of equipping contact-centre agents with an AI-based conversation assistant that offered real-time suggestions for responses. The agents remained in control of the conversation, and were able to accept or ignore the AI’s suggestions as they saw fit. The authors found that the tool increased worker productivity by 14%, measured by the number of chats agents were able to successfully resolve per hour. It also disproportionately improved the productivity of less experienced agents, meaning a more consistent service for customers.
A boost to productivity was not the only benefit wrought by AI. Mr Brynjolfsson and colleagues also measured the sentiment of customer interactions, and found that use of the technology led to a significant improvement in how agents were treated by customers, thanks perhaps to speedier resolution of their problems. They studied, too, the tool’s impact on the attrition rate of agents, which also improved.
For ever by your side
Since ChatGPT dazzled the world last November, many have fretted over whether AI will obliterate entire categories of jobs, not least among them contact-centre agents. The evidence so far hints instead at another, more hopeful possibility: by augmenting workers, rather than replacing them altogether, generative AI could lead both to better jobs and better experiences for customers. After years of frustration and rage, that would come as a relief to people on both sides of the customer-service line. ■
Read more from Schumpeter, our columnist on global business:
What Arm and Instacart say about the coming IPO wave (Sep 21st)
The Mittelstand will redeem German innovation (Sep 14th)
America’s bosses just won’t quit. That could spell trouble (Sep 4th)