Ask ChatGPT whether artificial intelligence will replace humans in customer service, and it will give you a diplomatic answer: that humans and machines will work side by side. Yet not everyone shares that optimism. The chief executive of Indian technology firm Tata Consultancy Services, K Krithivasan, told the Financial Times last year that AI may soon mean there is “minimal need” for call centres in Asia. Research firm Gartner echoes this outlook, predicting that by 2029, AI will autonomously resolve 80% of common customer service issues.
At the heart of this shift is the rise of so-called “AI agents” – autonomous systems capable of making their own decisions. These are designed to replace today’s rule-based chatbots, which can only answer from a fixed list of questions. For now, many companies still rely on older systems. When I contacted parcel delivery firm Evri recently, its chatbot Ezra promised to “get this resolved straight away.” Yet after confirming my parcel had been delivered – complete with a photo of a package left at the wrong door – the conversation simply stopped, with no option to continue. Evri says it is investing £57 million to improve its service, and that most customers get answers “within seconds.”
Other firms have found that giving AI more freedom can backfire. Rival delivery firm DPD was forced to disable its AI chatbot after it began criticising the company and swearing at customers. Finding the balance between personality, professionalism, and genuine problem-solving is proving a delicate task. According to Gartner, 85% of customer service leaders are now exploring or deploying AI chatbots, but only around 20% say their projects are fully meeting expectations.
“There’s no question that AI allows a much more natural conversation,” says Gartner analyst Emily Potosky. “But it can also hallucinate, give outdated information, or just be wrong.” She notes that for tightly defined tasks like parcel delivery, simple rule-based agents might still be the safest bet. While cost-cutting remains a major motivation, Potosky warns that AI isn’t always cheaper. “This is a very expensive technology,” she says. “And it depends heavily on high-quality training data and good knowledge management.”
That data often comes from existing call centres. Joe Inzerillo, chief digital officer at Salesforce, says these operations provide a rich training ground for AI. “You have a huge amount of documentation, and that’s really great material for AI to learn from,” he explains. Salesforce’s AI-powered platform, AgentForce, is already used by clients such as Formula 1, Prudential, Open Table, and Reddit. The company discovered that small human touches make a big difference – teaching its agents to say “sorry to hear that,” for instance, made interactions feel more authentic.
Salesforce also learned the importance of flexibility. When its AI refused to answer questions about integrating Microsoft Teams because the software appeared on a competitor list, customers were left frustrated. The rule was quickly removed. Today, Salesforce claims 94% of users choose to interact with its AI agents when given the option. The company says customer satisfaction now exceeds that achieved with human agents, and that AI has already helped it cut service costs by $100 million. However, Inzerillo stresses that this hasn’t meant mass layoffs – most affected staff have been redeployed elsewhere in the business.
For some, though, human connection remains irreplaceable. Fiona Coleman, who runs workforce technology firm QStory – used by clients like eBay and NatWest – believes AI can enhance working conditions for call centre staff but not replace them. “There are times when I just want to speak to a human,” she says. “Let’s see what happens in five years’ time – whether an AI can really handle something as sensitive as a mortgage application or a debt problem.” Her caution may be well placed: proposed U.S. legislation would require companies to disclose when AI is being used and allow customers to request a human at any time, while Gartner predicts that by 2028 the EU may enshrine a “right to talk to a human” in consumer law.
In the end, AI may revolutionise customer service – but empathy, trust, and the human touch remain hard to automate. The question is not just how capable AI becomes, but whether customers will ever truly want to give up the reassurance of speaking to a real person.
Derived from an article written by Jane Wakefield, Technology reporter for the BBC
ByJane WakefieldTechnology reporter
