Albany Times Union/Hearst Newspapers | Hearst Newspapers | Getty Images
Artificial intelligence may be the future of customer service, but some early consumer reviews suggest you should be prepared to be frustrated, at least for now.
AI-powered chatbots act as virtual concierges that guide wayward customers to the right solution, but many customer service chatbots still redirect rather than solve problems. Rejecting a request outright or sending the customer into a maze of AI-driven ambiguity where they can’t continue with their complaint are still common in chatbot playbooks.
“I hate AI customer service chatbots,” said Carmen Smith of Campo, Calif., who says the technology often sends her into an infinite loop. “In any case, they all seem to either present some kind of FAQ list or repeat information that they’ve already tried and felt is missing,” Smith said. “I hate dealing with them, but unfortunately a lot of companies do deal with them these days. I’d rather talk to a human.”
Smith is not alone. According to Qualtrics’ 2026 Customer Experience Trends Report, nearly one in five consumers who used AI for customer service saw no benefit from the experience. With a failure rate nearly four times higher than typical AI usage, this number points to something specific to customer service that makes it difficult for AI to respond well. Consumers rank AI applications for customer service as the worst in terms of convenience, time savings, and usefulness. “Too many companies are deploying AI to cut costs rather than solve problems, and customers can tell the difference,” said Isabel Zdatny, director of thought leadership at Qualtrics XM Institute and author of the report.
There are simple business reasons why many customers’ experiences are not positive. “AI doesn’t change incentives for companies; it expands them,” said Ben Weiner, global head of Cognizant Moment, the global technology and consulting firm’s digital experience business.
“Continuous optimization”
Companies have always shaped customer service around what they measure and what they reward. In many customer contact centers, human agents operate within tightly scripted flows designed to limit discretion. In other cases, brands are empowering their employees to do what they need to do to satisfy customers.
“If management prioritizes minimizing refunds, reducing human escalations, and shortening call times, you can expect an AI agent to reflect that philosophy in the experience, just as you would a human agent. Those are always business choices, and AI systems will enforce those choices as well,” Wiener said, adding that AI can do that more consistently and at high volume. “AI relentlessly optimizes every metric given to it,” Wiener says. “Companies need to be clear about what outcomes they want their AI systems to prioritize, because they will deliver exactly what they are trained and measured to achieve,” he added.
“What bothers them is automation that keeps them in a loop,” said Shannon McKean, professor of practice and executive director of the Center for Analytics Impact at Wake Forest University School of Business. Research on support automation shows that many conversations with AI still end up escalating to a human. But when a system can’t solve a problem or clearly explain a decision, customers often experience the AI layer as an additional barrier rather than a solution, McKean said.
Deflection has benefits for humans working in customer service.
According to Terra Higginson, principal research director at Info-Tech Research Group, deviations in AI are justified when it is used to protect workers in jobs with high burnout, turnover, and mental health issues.
And in some cases, saying no is the right decision.
“If two people are arguing about a refund, and the law says there can’t be a refund, the judge will rule rather than argue endlessly, which is often the case in scenarios of an agent and a disgruntled customer,” Higginson said. “This makes the process of enforcing rules and regulations rather than refunds difficult,” Higginson said, adding that AI can consistently enforce rules across the board in a way that humans can’t, “without the stress of following internal rules and getting yelled at.”
On the other hand, having difficulty securing a legitimate refund is just bad business, and always has been. “That’s obstruction, not service,” Higginson said. Higginson added that this is a particularly bad business model in a competitive market where digital opinions can spread quickly through forums and social media.
Consumer chatbots are here to stay
Tom Eggemeier, CEO of Zendesk, says that too many companies define “resolved” interactions in a way that includes response deviations and non-responses. Zendesk only counts a resolution if the customer, company, and employees all agree that the problem is actually resolved. “AI is a means, not an end,” Eggemeyer said.
One solution he thinks is likely to come to fruition in the not-too-distant future. The idea is for consumers to have personal AI agents to deal with corporate chatbots, allowing AI to solve low-level problems.
Consumers may need help.
Eggemeyer predicts that within three years, 50 percent of digital customer service interactions will be handled by AI, rising to 80 percent within five years.
Jesse Chan, CEO of customer service chatbot developer Decagon, which has closed more than 100 business deals across consumer industries in 2025 and tripled its valuation to $4.5 billion in a recent funding round, says companies that try to deflect customers will lose money in the long run.
“I haven’t met a single customer who intended to change direction,” Zhang said. “People are very active in optimizing resolution,” he added.
Sierra, a conversational AI platform founded in 2023 by former Salesforce co-CEO Brett Taylor and former Google executive Clay Baber, says its business model employs “outcome-based pricing” and believes it is a key way to approach these new interactions. “If we’re not solving a problem, if it’s not working for our customers, it’s not working for us,” a Sierra spokesperson said.
Zhang acknowledged that from the customer’s perspective, subjectivity can creep in on this topic, and one person’s resolution is another person’s disagreement. But he said it’s up to companies to deploy AI that’s smart enough to make decisions. “You can’t say no to everything, you can’t say yes to everything. We need a solution,” Chan said.
What you don’t want is a dead end, but an “escalation path” for customers who don’t get what they need from the AI answer.
There are situations where there must always be a clear and fast path to a human agent, such as elderly customers, VIP customers, or particularly complex issues.
A widely cited example of an AI chatbot implementation is fintech Klarna. While AI is not the only factor here, it played a key role in the recent 40% headcount reduction. However, in an AI-first shift, the company ultimately decided to rehire some employees in its customer service department due to poor performance from AI technology on more complex tasks. A company spokesperson recently told CNBC in a statement that Klarna remains focused on using AI, launching an AI assistant that took on the jobs of 700 customer service agents at launch, which has now grown to 800 agents. The AI assistant is now accepting more customer inquiries and its customer satisfaction scores are on par with human agents, the spokesperson said.
The market is rapidly evolving, which can lead to different customer experiences. “Consumers sometimes don’t know the difference between traditional chatbots and AI. Traditional chatbots can’t do things or solve problems,” said a Sierra spokesperson. Some brands are being “very cautious” about implementing new AI chatbot technology, putting so many guardrails on their AI agents that they can’t even make the decisions needed to solve a problem. “They do it because they’re afraid of making a mistake, but the guardrails have to be reasonable,” she says.
More complex AI use cases also exist at a sectoral level, such as in healthcare. At NotifyMD, AI is used in customer service to address some of the simpler issues, such as answering customer calls regarding billing. But Jody Miller, senior vice president of sales, said humans are still important when it comes to something complex and emotional. “There is simply no way that AI can bring the kind of understanding and empathy that a human can bring, especially when a customer is upset or has a legitimate problem,” Miller said. “Going forward, I think the key for all of these companies is to be very careful about how they use AI and to make sure that it helps rather than hinders the people who need it most,” Miller added.
Zhang believes that the future of customer service will be AI, and AI agents will have memory and be able to respond to all types of customer service inquiries. “At a very high level, every company will have AI on the front end and one unified agent across all channels,” he said.
