How Soon Will AI Replace Customer Service Agents?


Share on LinkedIn

A recent article by Pranshu Verma in the Washington Post argues that contact centers are toast. The article features examples of contact center operations that have been completely replaced by Generative Artificial Intelligence (GenAI) systems, such as ChatGPT.

The Washington Post quotes Sumit Shah, founder of e-commerce platform Dukaan, saying: “It was [a] no-brainer for me to replace the entire team with a bot, which is like 100 times smarter, who is instant, and who cost me like 100th of what I used to pay to the support team.”

Commentators have been suggesting that professional jobs are about to vanish since ChatGPT was released at the end of 2022. The Guardian ran a feature in February that also suggested contact centers will be one of the first types of job to vanish.

But the point to be aware of is that AI is good at performing tasks. It still can’t mimic the entirety of functions a human is capable of. It replaces tasks, not entire jobs. Change is coming, but in the short term the most noticeable impact to customer service will be dramatically improved chatbots for the customer and much better internal systems inside the contact center – agents using AI to perform most of their admin tasks.

Dukaan is a specific type of service. It’s a platform that allows anyone to create an e-commerce store without coding. So it is app-based and therefore much more likely to require in-app text and messaging support rather than a real person on a phone call.

There are still several reasons why it can be advantageous to retain real humans in the customer service process, for example:

  1. The support channel: As mentioned, in-app support works particularly well with text support, but there are many occasions where a customer wants to hear a real person that can help immediately – such as when a flight is delayed or when they want to talk about a financial transaction.
  2. The value of human interaction: There can be opportunities to offer ideas, recommendations, and advice to the customer that leads to upsell and cross-sell opportunities. For example, if an online customer has filled their shopping cart, but is delaying the checkout process then a personal interaction to ask if they need help can be all it takes to encourage them to confirm payment, rather than abandoning the cart.
  3. Novel interactions: AI is smart, but it needs to be trained. Sometimes there will be new problems that the AI has never seen, maybe a problem the entire customer service team has never seen, the type of unusual problem that needs investigation and analysis by an expert.

Nobody can deny that the customer journey is changing. A modern customer with a problem or question is likely to ask Google for help, or check YouTube for an explainer video, or ask Alexa, or interact with the support chatbot. All these steps are extremely likely before the customer uses a support phone or chat service.

This means that the way customer service has to be designed is also evolving. By the time the customer does interact with an agent, they may have tried two or three self-service options. Your agent needs to be better than Google and the AI-powered chatbot. The days of handing a customer service agent a script are long gone.

So here are three likely outcomes of the AI revolution, beyond just stating ‘contact centers are toast.’ These are the questions that major news organizations such as the Washington Post and Guardian should really be exploring.

The role of customer service agents will evolve

Forget the minimum wage entry-level jobs of the past, if the AI and improved self-service is handling most interactions then the remaining human interactions will be problems that need smart investigators – real experts. This elevates the status of the customer service role – agents become subject matter experts. It is likely these will be home-based experts, rather than armies of people based in a single contact center, because if you need experts in League of Legends to provide in-game support then spreading the net globally is more realistic than hoping enough experts live near your Belfast contact center.

Fewer agents, smarter, better paid, in more interesting jobs.

The way companies pay for customer service will evolve

Customer service is typically paid for by the FTE – Full-Time Equivalent person in a contact center. Also known as bums on seats. If the volume of customer interactions requires 200 people in a contact center then the service is charged as 200 x $X… It’s simple and easy to plan for, but if the customer journey uses more sophisticated self-service elements and fewer interactions with agents then the model starts breaking apart. Who is paying for the AI if the traditional charge has only ever been for people in a contact center?

We will need to start seeing contracts that charge for customer satisfaction or transformation targets. It’s harder to measure, but this will need to change.

Total lifetime value

Customer service used to be all about calling after a purchase. I have a complaint. I have a problem. I need help. I can’t get my new TV to connect to the internet. It was transactional and revolved around fixing customer problems. Now customers build a relationship with their favorite brands. Interactions don’t need to be transactional, they can just be about reinforcing the relationship. If this sounds odd then look at how fans of bakery chain Greggs interact with the company. Fans of Nike have tagged Instagram photos #Nike over 118 million times. Nine million people follow Land Rover just to see the latest photos of their favorite vehicle.

The lifetime value of customers and the relationship that is created is far more valuable than any individual transaction – customer service has to be designed to nurture these relationships, not just get a customer off the phone quickly.

It is not just AI that is dramatically changing how customer service is designed and how contact centers will function in future – the way that brands interact with customers is evolving and AI is becoming an important tool that is accelerating the changes that were already taking place.

Let me know what you think about the bots, AI, and customer service. Change is coming, but how do you think it will look in 2024 and beyond? Leave a comment here on this page or contact me direct on LinkedIn.

Mark Hillary
Mark Hillary is a British analyst and writer based in São Paulo, Brazil. He has written 25 books on CX and technology and he hosts the CX Files podcast with Canadian CX analyst Peter Ryan.


  1. If ChatGPT ever replaces live customer service agents, both the customer and the organization using this technology will be the poorer for it. Customers are already super-frustrated with the ability of robotic, automatic electronic service to efficiently and effectively resolve their issues. ChatGPT, from all the demos I’ve seen, has the potential to take this frustration, and resulting customer negativity, to another level. Humanity must be built into service operations.

  2. This is an excellent piece, well researched and thoughtful. The bottomline is: What do customers want (not just what they expect). When express check-in happened in major hotels, a lot of guests were happy they could completely by-pass the front desk and go straight to their guest room. But, some guests still wanted to shape room choice and build a relationship (potentially useful later) with the human at the front desk. Smart hotels provided both. AI will change the contact center, not eliminate humans. Here are three reasons.

    AI Has No Heart

    As captivating as Wall-E and ET were to moviegoers, the AI inside the characters had no heart. According to researchers, even after significant advances, AI is still very far from being able to feel emotions in the same way humans do. “They can recognize and respond to emotions, but they do not have the same complex psychological and physiological responses humans do,” says software engineer Bimo Tristianto.

    AI Has No Instinct

    “Being customer-focused, not competitor-focused,” says Amazon founder Jeff Bezos, “allows you to be more pioneering.” Future perfect thinking requires a deep understanding of customers and an intuitive leap of faith in what they want or need. The two astronauts in the 1968 hit movie 2001: A Space Odyssey defeated a rogue computer named Hal, not by doing what they had been trained to do, but through an innovative solution that defied conventional machine logic. According to research, ChatGPT cannot intuit. Its responses vary widely in reaction to tiny differences in how questions are phrased.

    AI Has No Soul

    “The customer is not always right,” said Stew Leonard, Jr., CEO of the famous grocery chain that bears his father’s name. “But it’s our job to make them feel right.” Acceptance and forgiveness are vital attributes in the customer loyalty world. Laura DeCarlo, of Career Directors International wrote, “When things go wrong, assume the client simply wasn’t paying attention and be prepared to guide them with kindness, compassion, and professionalism.” AI service is wired to be accurate but not to be empathetic. When Google engineer Blake Lemoine asked Google executives whether a chatbot had a soul, he said the idea was scorned. “I was literally laughed at by one of the vice presidents who said, ‘Oh, souls aren’t the kind of things we take seriously at Google.'”

    AI, even in its most advanced form, is not and will not become a junior version of human. In the words of neurologist Oliver Sacks, “The correlates of consciousness root it in the life of the body, the pulse-beat of experience hungry for meaning–something lacking in a machine of even the most astonishing computational capacity.”

    The world of the contact center is rapidly changing. And, AI will play a vital role. However, when mixing AI with customer relationships, never forget what the “A” stands for. Without the capacity for authentic personalization as a human, it will best succeed in the “backroom” with coders, researchers, and auditors, leaving the front of the house for people who can remember to serve customers like a genuine, caring friend. Again, the bottomline is a question: What does the customer want?

  3. Hi Mark: thank you for this article. My first observation is that a different question – actually questions – need to be asked: 1) How will customer service expectations evolve? and 2) which roles will AI perform in fulfilling or supporting those evolving expectations?

    A quibble: in too many instances, I have seen a nascent information technology deployed in business, immediately followed by a chorus of hyperbole. “[Name of nascent technology] will replace [name of legacy technology]!” This binary view is not only worn out, it’s been repeatedly disproven. There are many business processes that have been supported or changed with new technology. Accordingly, I find replacement more marketing hype than reality. The Internet was available for customer support about four decades ago, but I still occasionally contact companies by phone.

    Establishing a clear understanding of what, exactly, needs to be solved is a necessary first step. The right questions will logically follow. Many times, companies simply start with “How can we use AI?” or “How can AI help us cut costs.” Companies that launch their AI effort with those interrogative starting points create considerable strategic wreckage. Your final section that covers total lifetime value hints that as an artifact of changing customer needs, customer service itself is changing, and AI will play a central role as it evolves.

  4. A well-framed perspective, Mark – and I agree that “the lifetime value of customers and the relationship that is created is far more valuable than any individual transaction.”

    When I had read the Post article, all I could think of was ‘here we go again.’ The answer to “How soon will AI replace agents” is: Immediately. I suspect it will be the same pattern as early-adopters of IVR and chatbots. They will do it poorly, but justify it with short-sighted economics.

    I have a vivid memory of a senior V.P. of a telecom giant explaining the economics of their feet-first plunge into a clunky, nasty IVR system. Like Mr. Sha’s argument, it centered on cost per interaction. We had been analyzing the company’s churn rates at the time. One year in, two weeks before they were going to announce the monstrous cost savings they had achieved, we showed them two numbers. The first was how much had been lost in increased customer churn (and the CLV potential loss). The second was the dramatic increase in calls-per-customer as frustrated customers had to reach out sometimes 3-4 times to actually get a human. They never made their announcement.

    Researchers long ago identified that the root of customer loyalty is trust, and that trust is created though caring, competency and integrity/ownership. AI will ultimately bring competency and may eventually bring a semblance of integrity and ownership. As Chip pointed out so well, however, it’s ability to care in the way that humans do – and act accordingly – doesn’t currently exist, and may never.

  5. looking at the title … isn’t it fair to say that it already does? The more efficient agents become the lower the number of agents needed?


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here