Everyone has an automated customer service horror story, but for many consumers, the horror is less about man versus machine than about getting a satisfying outcome from a service call.
A recent survey of 3,000 consumers by Gradient Labs, which develops AI agents for financial services customer operations, found that 49% would prefer an AI agent to fix their problem immediately rather than wait until the next day to speak to a human.
Among respondents who had used an AI agent for a customer service issue in the past year, 87% reported that their issue was fully or partially resolved.
But the survey also revealed that four out of five respondents have deliberately tried to circumvent an AI agent to reach a human customer service representative.
Conflicting findings? Not necessarily.
Contacting a human gives you someone to talk to, but it doesn’t guarantee an outcome, Gradient co-founder and CTO Neal Lathia explained.
“Most chatbots people deal with can answer a question, as they typically aren’t able to reopen a dispute or apply a policy,” Lathia said in a statement. “This is what drives people to ask for a human.”
“The demand changes once you give the AI the ability to actually finish the job,” he continued.
“Interestingly enough,” he added, “an AI agent who truly understands the domain will know when it is out of its depth and, consequently, the moment when a person should step in.”
Limited AI Access Can Frustrate Customers
AI agents aren’t given enough agency to solve real problems, and analysts reinforce that by telling companies to “take it slow” with “baby steps” and start with a simple FAQ AI agent so they can test it and see how it works, said Jen Grant, CMO of Quiq, a maker of an AI-powered customer experience platform in Bozeman, Mont.
“But we have not seen that to be successful at Quiq,” she told TechNewsWorld.
“Human agents are given access to order management, customer records, or other critical systems within the business, so they can look up, solve, and update records for the customer,” she explained.
“But if the AI agents, whether on voice, text, or chat, don’t have the same access to systems that the human agents have,” she continued, “they can’t solve many of the real problems customers have, and so it has to be escalated to a human.”
“That’s where the real problem starts,” she said. “Businesses have, in their effort to be cautious and ‘start small,’ put AI agents into the customer support flow that can’t do too much and then turn into yet another hurdle for their customers to get an answer.”
“Whether you want to reschedule a technician visit, change the dates of that room you booked, or just want information about the return policy for the thing you ordered, if you still have to get transferred from an AI agent to a human agent, that makes the experience even longer, more frustrating and worse than it was before AI was introduced,” she noted.
“Consumers start to see the AI agent as a frustrating roadblock, and that’s why laws start to get put into place,” she added.
Right to Human Customer Service Act
California has passed one such law. The Right to Human Customer Service Act, AB 1609, takes effect Jan. 1, 2027. Subject to exemptions, it covers businesses with more than $500 million in national gross annual revenue and requires a simple way to request human assistance during regular business hours.
Covered businesses must make a good-faith effort to connect customers with a human agent within 15 minutes of a request or provide a specific appointment within one business day for an agent to call or connect with them.
In addition, covered businesses cannot represent a customer service chatbot as human. They must clearly disclose its artificial identity when a reasonable person would likely mistake it for a human.
“A law is needed because some companies have made it too difficult for customers to reach someone who can resolve their problem when automated support fails,” argued Brett Faulk, senior vice president and principal analyst at SmartTech Research, a technology advisory firm in Las Vegas.
“Consumers deserve a clear, timely path to help instead of repeated chatbot exchanges and long hold times,” he told TechNewsWorld.
“The law should help customers escape automated loops, but reaching a human does not guarantee resolution,” he added. “If that representative lacks the authority, information or tools to fix the problem, consumers simply encounter the same obstacle through a different channel.”
When Automation Becomes a Barrier
Hitee Chandra Jha, co-inventor of a patent on semi-automated form-based chat, pointed out that when automation is measured primarily by avoided human contact rather than successful resolution, it can become a barrier instead of a service.
“A law can establish a minimum right to recourse for customers who are otherwise stuck inside an automated flow they cannot redirect,” she told TechNewsWorld.
“Many companies have used automation mainly to cut costs and reduce contact with customers, rather than to solve problems faster,” added Nate MacLeitch, founder and CEO of QuickBlox, a global builder of AI agents, chat and video tools that businesses use for customer and patient communication.
“When there’s no clear way out of a chatbot or phone menu, people feel trapped, and that’s what this law is responding to,” he told TechNewsWorld.
MacLeitch said QuickBlox’s experience showed why customers need a clear handoff from AI to human support. “We learned why the hard way,” he said.
“Our first AI product was a knowledge AI agent, which could be trained on a business’s own information to answer customer questions,” he continued. “The feedback from potential customers was that answering questions wasn’t enough. Some conversations will always need a person, and what happens at that point matters most.”
“So we built an AI agent workflow platform to manage the handoff,” he continued. “The conversation with the AI isn’t lost, the customer is routed to the right person, and that person sees what has already been said.”
“A legal right to reach a human helps, but if the human starts from scratch and the customer has to explain everything again, the frustration just moves to a different place,” he explained.
Feeding Anti-AI Bias
Ideally, companies would design their systems so when they can’t resolve a problem, consumers should have a reasonable opportunity to reach someone who can, noted Anthony Miyazaki, a professor of marketing at Florida International University in Miami.
“When companies fail to do so, government intervention becomes more understandable,” he told TechNewsWorld. “The issue is whether the government intervention will truly resolve the problem or even create new ones.”
“Giving customers access to a human representative is not the same as giving them access to a solution,” he explained. “If companies respond to the law by staffing customer service operations with inexpensive but inadequately trained representatives, they may simply replace frustrating automated experiences with frustrating human ones.”
Miyazaki argued that chatbot disclosures required under the California law could unintentionally bias customers’ evaluations of the service they receive.
“The effectiveness of customer service should be judged by the quality of the solution, not by whether the provider is artificial intelligence or a human representative,” he said.
“While transparency about AI use is important in certain circumstances, requiring companies to disclose whether customers are interacting with AI or a human may activate biases that negatively influence their perceptions of the service they receive, even when the underlying service is identical,” he continued.
“Companies routinely serve customers through representatives located in other countries without proactively disclosing their locations,” he observed, “so perhaps we should be asking whether the identity of the service provider matters more than the quality and effectiveness of the service itself.”
