NEW YORK (AP) — Armen Kirakosian remembers the frustration of his first job as a call center agent nearly a decade ago. Deteriorating customers, constantly searching through menus, remembering notes that he needs to physically write about each call he handled.
Thanks to artificial intelligence, the 29-year-old from Athens, Greece, has not written notes or clicked countless menus. He often has a full customer profile before him when a person makes a call, and may already know what problems the customer has before saying “hello.” He can actually spend more time with his customers.
“AI took the robot out of us,” Kirakosian said.
Approximately 3 million Americans work in call center jobs, millions more work in call centers around the world, answering billions of inquiries a year, from broken iPhones to shoe orders. Kirakosian works for companies that provide third-party customer service lines from 22 countries to companies in the industry, such as automobiles and banks, that require additional capabilities or outsource the operation of call centers.
Answering these calls can be a job that is not grateful for. According to McKinsey, about half of all customer service agents quit their jobs after a year, and stress and monotonous jobs are one of the reasons employees quit.
Many of the things these agents deal with are referred to in the industry as “breaks/fixes.” This means something is broken, wrong, or confused. Customers expect the phone person to fix the issue. Now it’s a question of who will be tasked with fixing. You are a human, a computer, or a human being enhanced by a computer.
Already, AI agents are taking over more routine call center tasks. Some jobs were lost and there were disastrous predictions about the future job market for these individuals. This leaves half of all call centre jobs over the next decade from a modest single pint point loss. However, as some customer service issues become increasingly difficult to solve, the industry has even higher levels of learning and training, and the decline may not coincide with more disastrous predictions as it becomes clear that the industry still needs humans.
Some financial companies have already experimented with moving heavily with AI on customer service issues.
Swedish shopping Klarna paid later to the company, replacing 700 of the roughly 3,000 customer service agents with chatbots and AI in 2024. The results were mixed. The company saved money, but Klarna discovered that in certain circumstances, more skilled human agents are still needed, including complex issues associated with identity theft. Earlier this year, Klarna hired seven internal freelancers to handle these issues.
Earlier this year, Klarna sent a small number of customer service employees back to the company and admitted there were certain issues that AI couldn’t handle like real people, like identity theft.
“The vision of the AI First Contact Center is quickly becoming a reality as AI agents handle most of the conversation and fewer, better-trained human agents support only the most complex tasks,” said Gadi Shamia, an AI software company that sounds more and more human, in an interview with a consultant at McKinsey.
The experience of call centre customers has improved, but it is still far from perfect.
The first customer service calls have long been handled through an interactive voice response system known in the industry as IVR. Customers interact with IVR when they are told “press one for sale, press two for support, press five for billing.” These crude oil systems received an update in the 2010s, when customers could encourage their systems by saying simple phrases like “selling” or “support” or “want to pay the bill” instead of navigating maze menu options.
However, customers have almost patience with these menus, leading them to “zero out.” This is call centre slang when a customer hits a zero button on the keypad in the hopes of reaching a human. It is also not uncommon for them to be held and transferred after the customer “zero” was “zero” as they were not in the right place for their requests.
IVR, recognized the collective impatience of Americans with Sen. Reuben Gallego, Democrat Sen., of Arizona, and Jim Justice, of Republican, of West Virginia. This introduced “maintaining American call centers,” which requires a clear way to contact human agents who maintain jobs in US call centers and provide incentives.
Companies are trying to deploy telephone systems that broadly understand customer service requests without navigating through menus and predict where customers will be sent. Openai, the maker of ChatGpt, offers a “ChatGpt Agent” service for users who can understand phrases such as “I need to find a hotel for my wedding next year.” Please tell me what clothes and gift options are.
Bank of America says it has successfully integrated such features into Erika, a chatbot that debuted in 2018. If Erika is unable to process the request, the agent will direct the customer to the appropriate department. Erica is now predictive and analytical, knowing that customers may be repeatedly underbalanced and need better support, for example.
Bank of America said earlier this month that it has been used by Erica three billion times since its creation, and customer service requests are increasing. The chatbot’s moniker comes from the last five letters of the company’s name.
James Bednar, Vice President of Products and Innovation at TTEC, has spent much of his career trying to reduce customer service calls from the company as well as the pain of the company. He said these tools could ultimately kill the IVR forever, and that anyone could end the need to “zero out.”
“We’re approaching the point where AI will bring you to the right person for your problems without you having to route those menus,” Bednar said.
