Automating Call Centers with AI

There’s good news and bad news for those whose blood pressure shoots up when a call center directs you to a chatbot without even the possibility of first conversing with a human.

First, the bad: the chances of speaking with a human customer service representative will continue to diminish. Companies are rapidly adopting artificial intelligence (AI) to support what is less likely to be referred to as a “call center” and now typically named a “customer contact center” that handles telephone calls plus all digital channels of inquiry, including messenger apps, email, texts, and website pop-up windows.

With the arrival of AI, the human headcount in the contact center—including people who might otherwise answer a call and talk with you—is shrinking.

The extent of this reduction is debatable, but its direction is not: call centers-cum-contact centers are replacing many workers with some aspect of AI.

“I’d be lying if I said it wasn’t replacing people,” said Josh Streets, technology practice leader of the International Customer Management Institute (ICMI), a London-based organization with the stated mission to “champion contact centers and their people.” Chicago-based Street’s job is to help contact centers with the uptake of AI and other digital tools. While this might lead to more effective contact center operations, it does not bode well for job security for conventional call center operators.

There are some voracious outlooks for AI replacing humans in call centers, albeit not completely replacing.

“I don’t believe in 100% automation, I believe in 80% automation,” said Matthias Goehler, the Mannheim, Germany-based EMEA chief technology officer for San Francisco customer connection technology company Zendesk.  “There will be humans, but their roles will change.”

Goehler’s notion of change is echoed by others in the industry.

“Do I see AI eventually replacing some jobs in the contact center? Yes, I do,” said Liz Miller, vice president and principal analyst with Cupertino, CA-based advisory firm Constellation Research. “It’s going to replace jobs that should be replaced. It’s happening. It’s not a huge wave. We’re not seeing headcount dropping or total replacements right now, but it is replacing those jobs that probably shouldn’t have existed if we were really being efficient in the first place.”

Getting Better All The Time

Therein lies what could be the good news: advocates of AI in the call center say AI will improve the customer experience.

“What we’re going to start to see is that, because of the efficiency, because some of the little things can just be handled through self-service engagement and interaction, we can start to reduce the headcount,” said Miller. “But we can start to invest in headcount that is really well-paid, really well-trained, that is really focused on very different aspects of the contact center. It will open up opportunities to invest in different capacities and capabilities of the contact center to invest in better experiences.”

Zendesk’s Goehler said that major improvements are happening now that AI has hit its “agentic” stage. Agentic AI, compared to earlier generations, is better able to work autonomously towards a goal—of understanding a customer’s specific situation, for instance—in part because of advances in underlying large language models that make it a nimbler, more flexible actor, rather than a device aimed at a narrow task.

In principle, agentic AI can handle gray areas that fall outside of the classic “press 1, press 2, press 3” prompts. For example, an appliance company’s contact center will be able to infer that a customer’s new refrigerator is not broken, but that it arrived absent a freezer flap. A towing company will know that a caller is not phoning about a car’s breakdown, but rather about an automobile recall.

“With agentic AI, the AI agent really can handle multiple intents and changing intents,” said Goehler. “Now, instead of ‘pressing three for everything else,’ customers can just say what they want; the system will keep asking questions until it understands what they want. It won’t say ‘I’m sorry, I don’t know what you mean.’ How many times have you been on a call where you waited for 20 minutes, and then you got handed over three times, then you needed to repeat yourself three times? How great of an experience was this? If you can take some of this away, this will be a better experience for customers.”

The proficiency of agentic AI is a big reason why Goehler foresees a high 80% human replacement in the contact center, a level that he believes will happen within a year or two among employees working in digital channels such as email, messaging, Web pop-ups, and chatbots. The road to 80% will be longer—about three to five years—among voice employees, as AI is in need of improvements in speech technology, including speech recognition and speech-to-text transcriptions, he said.

Yet whether AI expressly replaces humans, its presence in contact centers is already indisputable. “It’s probably at 100% of all call center implementations,” Goehler said, rattling off the myriad functions it is providing, such as interactive voice response, call transcriptions, creating contextual information for human agents, co-piloting human agents by suggesting next actions, and managing workforces, among others. AI in chatbots, and especially the newer agentic AI, makes for a more flexible experience than more conventional non-AI, rules-based chatbots, he said.

All this adoption, in turn, will free up contact center employees to handle more sophisticated and strategic jobs.

“What AI in the contact center is doing today very well is removing the mundane repeatable tasks that can be handled by AI and offering self-service opportunities for both employees and for customers,” said Miller. “Things like password resets, checking account balance, set up an appointment; the easy stuff. That doesn’t need to be on human agents’ plates,” she said, claiming that “customers are like, ‘You know what, I don’t need to talk to a person for that, I’m happy to have a computer do it’.”

Is It Really Better?

While some customers might be happy with AI replacing contact center agents, others might not. There’s no denying that AI, for many, still triggers uneasiness.

“There has been a backlash among people,” said Miller. “We love to give technology human attributes, and when you think about it, what is the human attribute that we have long associated AI with since the 1970s? It’s that they are murderous villains. For a lot of people, the introduction to AI was HAL opening the pod bay doors and killing Frank” in the 1968 film 2001: A Space Odyssey.

A 2024 survey by Gartner Group, a Stamford, CT, research firm, found that 64% of customers would prefer companies not use AI for customer service. They cited the difficulty of reaching a person as their top concern.

The public perception of AI-based customer service is not helped by high-profile debacles, such as the 2024 tribunal ruling in British Columbia that ordered Air Canada to compensate a customer after a chatbot had misinformed him about bereavement-related air fares in 2022.

Nonetheless, the AI tide continues to rise. Miller noted that all the leading providers of contact center software are adding AI to their mix, including Menlo Park, CA-based Genesys; Ra’anana, Israel-based NiCE; San Ramon, CA-based Five9; Seattle-based Amazon Connect; Palo Alto, CA-based Talkdesk. Many of these train AI-based software using data they’ve collected from customers on top of notable AI models such as Gemini, Anthropic, ChatGPT, and the like. In principle this enables commercial providers to better support customers in a manner that reinforces the tone and image of the commercial brand, Miller said.

In its The State of the Contact Center 2024 report published last December, ICMI noted, “This year finds contact centers in the crosshairs of change. Adapting to meet customer expectations, opening support for new channels, accommodating a hybrid workforce, and laying the groundwork for AI tools are some of the major forces shaping this moment.”

The report is based on broad, in-depth surveys about trends and attitudes across the contact center industry. One anonymous respondent seemed to speak for many, saying: “We need to streamline work for agents and allow the system to handle routine activities, allowing agents to be freed up for more complex work.”

In other words, even people in the contact center industry are advocating for AI, potentially at the expense of their own jobs.

That means the days of talking to a friendly human (or even a grumpy one) on the helpline are rapidly dwindling.

Mark Halper is a freelance journalist based near Bristol, England. He covers everything from media moguls to subatomic particles.