In an article published in WIRED magazine, journalist Dillon Thompson describes how corporations' growing reliance on AI-based customer service chatbots is not streamlining service, but rather making the experience exhausting and cumbersome for consumers. Thompson shares his personal struggle to locate a missing electric bicycle worth approximately $2,000. His attempts to resolve the issue turned into a months-long journey through digital systems and virtual waiting rooms run by chatbots, without any way to reach a human representative who could help him solve the problem.
The Lost Shipment and the Mysterious Signature in the Kitchen
It all began when Thompson and his partner decided to purchase two electric bicycles after receiving bonuses at their respective workplaces. His partner’s e-bike, which was rich in advanced features, arrived at their home in Atlanta—an area characterized by many hills—just days after ordering. However, Thompson's bicycle, which was ordered separately from a different retailer, was delayed repeatedly.
One Wednesday evening, Thompson received a text message from the delivery company FedEx confirming that the shipment had been delivered to his address and signed for by the recipient. This message seemed impossible to him, since at that exact moment he was standing in his kitchen, bicycle-less, and busy air-frying chicken thighs. When he checked outside his apartment, he discovered that the package was not there. Checking the order confirmation revealed that the shipment's receipt was signed by someone with the mysterious initials "M.M.", which did not match his initials, his partner's, or those of any other resident in their building. The next morning, Thompson began his attempts to contact FedEx customer service, thereby starting what he describes as a modern, exhausting version of digital customer service hell.
The Digital "Sludge" Phenomenon and Human Labor Reduction
The frustration experienced by Thompson is not accidental, but rather reflects a broader trend in the modern business world. In recent years, corporations have been extensively implementing artificial intelligence in their customer service departments, often at the expense of employing human workers. In a survey of customer service managers published in April, 31 percent of respondents indicated that they had already reduced or were planning to reduce headcount due to the adoption of AI technologies. While the majority of managers surveyed claimed that they shift human service representatives to other roles or add new tasks to their workload instead of laying them off, other executives have spoken about the issue more directly and openly.
Verizon CEO Dan Schulman recently noted in an interview with Bloomberg that artificial intelligence could replace a "large percentage" of the company's customer service work, emphasizing that this is one of the business sectors most exposed to the changes brought about by the technology.
For consumers, these changes lead to a less human experience characterized by prolonged wait times and a lack of real answers—features that have characterized bad customer service for decades, but are now receiving renewed force. In some cases, these systems are intentionally operated using a tactic known in the industry as "sludge"—creating obstacles and friction in the service process in order to discourage customers who are seeking a resolution or a financial refund. Professor Ryan Hamilton, a researcher of consumer psychology and marketing at Emory University, explains that while "sludge" existed even before the advent of artificial intelligence, AI has amplified the dystopian nature of the phenomenon.
Consumers themselves are not satisfied with this trend. In a report published in May, which surveyed consumers in the United States, the United Kingdom, and Canada, 59 percent of respondents expressed frustration with AI-based customer service agents, while 85 percent indicated that they prefer to speak with a real human representative.
The Winding Journey With Banks, Police, and the Delivery Company
During Thompson's attempts to locate the lost bicycle, he encountered chatbots in almost every channel he turned to. FedEx's AI agents repeatedly ignored his requests to transfer to a human representative on the phone. Even the Atlanta Police Department made handling the incident cold and alienated; when he called to file a report on lost property, he was asked to leave his details with a chatbot and wait for an officer to get back to him.
Thompson tried to submit two police reports: one over the phone and one via the police department's website. The first time, no representative got back to him. The second time, he received a callback, but it happened during a work meeting and he missed it. No voicemail was left for him, and when he tried to call back the number that dialed, he found himself once again in the same digital waiting room managed by a chatbot.
At the same time, he continued to act opposite the business entities. He managed to convince FedEx to open an inquiry claim regarding the lost shipment, but the company closed the matter with an automated email stating that the package was indeed lost, but for the purpose of receiving compensation, he must contact the retailer that shipped the product. When he contacted the bicycle company, he managed to bypass their chatbot and speak on the phone with a human representative. However, the maximum compensation they were able to obtain for him from FedEx was a refund of shipping costs only—a meager amount representing only about one-tenth of the price he paid for the bicycle (approximately $200 out of about $2,000).
Following this, he tried to turn to his bank and credit card company to dispute the transaction, but both attempts led him down long digital rabbit holes filled with chatbots, at the end of which a single human representative informed him that they could not help him because the package was technically lost under the responsibility of the delivery company FedEx. Three months after the start of the affair, Thompson remained without the bicycle and with a financial loss of approximately $1,700.
Agentic AI and the Sunk-Cost Fallacy
Despite the immense interest and publicity that agentic AI is receiving, the technology is still in its infancy. Meta's Toolformer model, which was considered a significant breakthrough at the time due to its ability to independently turn to external software tools when encountering uncertainty, was developed only about three years ago.
Hamilton explains that many companies adopt chatbots and AI-based service systems without deeply understanding how much they harm the customer experience, or they are aware of the damage but are willing to accept this trade-off due to budget savings considerations. While companies in certain industries can afford poor customer service without severe damage to their businesses, companies in other industries cannot do so. In an era where computer interaction is becoming the standard, executives may be forced to find new ways to differentiate themselves, perhaps by providing better human customer service. There is a risk that the technology will "flatten the service dimension," such that every company in every industry will operate the exact same AI-based customer service center.
Professor Ravi Dhar, a lecturer at Yale University and director of the school's Center for Customer Insights, identifies the "sunk-cost fallacy" as a primary driver behind AI implementation in various business sectors, and not just in customer service. Global spending on AI tools is expected to rise sharply this year, leaving many managers deeply committed to implementing the technology even when things do not go as planned. "If you are a CEO, you get questions from all investors and Wall Street, such as: 'What is your AI strategy, and is it showing a return on investment? After all, you are spending so much money on this,'" Dhar explains. Hamilton adds that many managers make decisions based on "optimism," assuming that the technology will develop and close the gaps or that the service experience will not be so bad, but in practice, they are liable to damage their reputations due to poor service.
The Response of the Atlanta Police Department and FedEx
In response to Thompson's inquiry, the Atlanta Police Department stated that an officer was indeed dispatched to his home on the day he missed the phone call, and promised to send an officer once again. At the same time, they confirmed that the person who called him on the phone was an operator and not an officer, and that returning a call to this number leads the caller back to a system managed by a chatbot of the police's emergency and non-emergency line.
FedEx stated in response: "While we leverage AI and digital tools to offer fast, convenient self-service for everyday questions, we recognize that complex situations require human care and deeper support. We use technology to empower our team members’ ability to deliver the 'Purple Promise.' We are continuously refining our processes so that customers can get the assistance they need seamlessly and swiftly."
Despite these statements, Thompson concludes that his prolonged encounter with the AI agents, which often felt like a deliberate barrier designed to prevent human interaction, completely exhausted him and left him without answers and without a solution to his financial loss.