AI reshapes call centers, but complex problems keep humans on the line

NEW YORK — Artificial intelligence is remaking how companies answer customer inquiries, easing routine tasks for agents and speeding routing decisions, but industry leaders and experiments show some problems still require trained human judgment.

Armen Kirakosian, a 29-year-old customer service agent who works for TTEC in Athens, said AI tools have removed much of the rote work that once defined his job. He no longer scribbles notes or hunts through multiple menus; he often sees full customer profiles before answering and can anticipate a caller’s likely issue. "A.I. has taken (the) robot out of us," Kirakosian said.

Roughly 3 million Americans work in call center jobs, with millions more worldwide handling billions of inquiries a year about products and services, from mobile phones to retail orders. Many of those interactions are what the industry calls "break/fix" — customers calling because something is wrong and expecting the person on the phone to solve it. AI has already taken over many routine tasks in those exchanges, performing functions from automated routing to generating prepopulated case notes.

The shift has produced cost savings and efficiency gains but also stirred concern about job losses. Forecasts range from modest single-percentage-point declines to projections that as many as half of call center jobs could vanish over the next decade. Analysts say the reality will be uneven: companies will automate repetitive work, while retaining and retraining humans for more complex or sensitive cases.

Some firms have tested heavy AI deployment. Swedish fintech firm Klarna replaced about 700 of roughly 3,000 customer service roles with chatbots and other AI last year. Company executives said the move reduced costs but left gaps for tricky situations such as fraud or identity-theft disputes; Klarna later rehired a small number of agents and added internal freelancers to handle those cases.

Gadi Shamia, chief executive of Replicant, an AI vendor that builds conversational agents, described an "AI-first contact center" model in which automated agents manage the bulk of conversations while a smaller corps of better-trained and better-paid humans handle complex problems. "Our vision of an AI-first contact center, where AI agents handle the majority of conversations and fewer, better trained and better paid human agents support only the most complex tasks, is quickly becoming a reality," Shamia told consultants at McKinsey.

Banks and major technology firms are among early adopters of predictive and conversational systems. Bank of America said its virtual assistant, Erica, introduced in 2018, has been used about 3 billion times and now offers predictive analytics — identifying customers who may need budgeting help or who have duplicate subscriptions — and can route calls directly to the appropriate human team when a request exceeds its capabilities. OpenAI has announced a "ChatGPT Agent" product intended to interpret complex customer instructions and coordinate multi-step tasks.

Despite those advances, legacy interactive voice response systems — those automated menus that tell callers to "press one for sales, press two for support" — remain a source of frustration for many consumers. Customers often "zero out" to reach a human, only to be put on hold or transferred because the caller was routed to the wrong department. Industry executives say conversational AI and better routing could largely eliminate that problem by matching a caller's intent to the proper team without manual menu navigation.

James Bednar, vice president of product and innovation at TTEC, said the technology is moving toward that result. "We're getting to the point where AI will get you to the right person for your problem without you having to route through those menus," he said.

The technology shift has drawn political attention. Senators Ruben Gallego, Democrat of Arizona, and Jim Justice, Republican of West Virginia, introduced the "Keep Call Centers in America Act," which would require clear ways for customers to reach a human agent and would offer incentives for companies that maintain domestic call center employment.

Workforce dynamics are changing as companies seek employees with higher technical and problem-solving skills to complement AI tools. Turnover in the industry has long been high; McKinsey and other analysts estimate roughly half of customer service agents leave within a year, citing monotony and stress. Employers and AI vendors say automation can reduce repetitive burdens and improve job quality for those agents who remain, but that requires investment in training and often higher compensation for staff who handle escalations.

Regulators, company managers and labor advocates are watching how the technology’s deployment unfolds. Companies cite efficiency gains and better customer experiences, while critics warn that automation without adequate human oversight can mishandle complex, high-stakes situations such as fraud, identity theft and nuanced billing disputes. The broad trend, industry executives say, is toward hybrid models where AI and humans collaborate: software handles predictable flows and data retrieval, while people focus on judgement, empathy and multi-step problem solving.

As firms refine how they deploy AI, the call center industry appears likely to change the composition and skills of its workforce rather than disappear. The speed of that transition and its effects on jobs and customer outcomes will depend on how companies balance automation with investments in human expertise and oversight.