AI transforms call centers, but some jobs still require a human touch


Armen Kirakosian recalls the challenges of his early days as a call center agent nearly a decade ago: upset callers, endless menu navigation, and manually jotting down notes after every call.
Now 29 and working in Athens, Greece, Kirakosian’s experience has changed dramatically thanks to artificial intelligence. He no longer has to take notes or endlessly click through systems. When a customer calls, their profile and likely issue are often already on his screen—allowing him to focus more on solving the problem.
“A.I. has taken the robot out of us,” he said.
Around 3 million Americans and millions more worldwide work in call centers, handling billions of customer inquiries each year—from tech support to online orders. Kirakosian is employed by TTEC, a global outsourcing firm that manages customer service operations for clients in sectors like banking and automotive.
Call center jobs are notoriously stressful and repetitive, contributing to high turnover rates—about half of agents quit within a year, according to McKinsey.
Many of these jobs involve "break/fix" scenarios—resolving issues with broken or malfunctioning products or services. Increasingly, the question is not whether the issue can be fixed, but whether the fix will come from a human, an AI agent, or a combination of both.
AI’s Growing Role in Call Centers
Artificial intelligence now handles many routine tasks in call centers. While some positions have been lost to automation, experts suggest predictions of mass job losses—up to 50% of roles—may be overstated. As customer issues become more complex, companies are still finding that human agents are essential, albeit more highly trained and specialized.
Klarna, the Swedish "buy now, pay later" company, cut 700 of its roughly 3,000 customer service jobs in 2024, replacing them with AI and chatbots. While the move saved money, it also revealed the limitations of AI. For example, cases involving identity theft still required human intervention. Klarna ended up rehiring a small team of experienced agents to handle those sensitive tasks.
“Our vision is that AI agents will handle most customer interactions, with fewer, better-trained, and better-paid human agents stepping in for the most complex issues,” said Gadi Shamia of Replicant, an AI firm that helps chatbots sound more natural.
Customers Still Face Frustrations
While AI has improved the efficiency of call centers, the customer experience isn't flawless. Most service calls still begin with interactive voice response (IVR) systems—the automated menus that ask callers to “press one for sales” or “press two for support.” These systems evolved in the 2010s to allow voice commands, but they remain frustrating for many.
Customers often “zero out”—hitting the “0” key repeatedly to bypass the system and reach a human. This sometimes leads to being placed on hold or sent to the wrong department.
Due to public dissatisfaction with IVR systems, U.S. lawmakers have introduced the “Keep Call Centers in America Act,” which would mandate clear paths to human agents and reward companies that keep support jobs domestic.
To address these frustrations, companies are now developing AI tools that can interpret full customer requests and route them accurately—without menus. OpenAI’s upcoming “ChatGPT Agent” is an example. It can understand complex phrases like, “I need a hotel for a wedding next year and suggestions for gifts and clothes.”
Real-World AI in Action
Bank of America has found success with “Erica,” its AI chatbot launched in 2018. Erica now handles more advanced queries and can predict customer needs—like helping someone with repeated low balances or managing duplicate subscriptions. When Erica can’t solve an issue, it seamlessly transfers the customer to the appropriate department.
According to the bank, Erica has been used more than 3 billion times and continues to take on a growing share of service tasks.
TTEC’s VP of Product and Innovation, James Bednar, says AI is on track to eliminate IVR altogether.
“We’re getting closer to a future where AI gets you to the right person without having to go through those frustrating menus,” Bednar said.