He Lost a $2,000 Ebike. Then the Chatbots Took Over.
A missing delivery turned into a months-long loop of automated dead ends. His story reflects a growing shift in customer service that is leaving real people stranded.
Key points
- 31 percent of customer service leaders surveyed in April 2025 said they have already reduced, or plan to reduce, staff headcount because of AI adoption.
- A May 2025 report found 59 percent of US, UK and Canadian consumers frustrated with AI customer service agents.
- In that same report, 85 percent of consumers said they would prefer to speak with a real human agent.
- Verizon CEO Dan Schulman told Bloomberg in 2025 that AI will likely replace a "large percentage" of the company's customer service work.
- One journalist spent nearly three months in chatbot loops after a $2,000 ebike went missing, recovering only about $200 in shipping fees.
The package arrived. Nobody in the building matched the initials on the signature, and the bike was gone.
That is where the trouble started for a writer at Wired, who splurged on a nearly $2,000 ebike and watched it vanish somewhere between a FedEx truck and his Atlanta apartment door. A text confirmed delivery. The kitchen was bicycle-free.
What followed was not just bad luck. It was a preview of something many shoppers are starting to hit.
Is AI customer service being used against you?
Yes, sometimes deliberately. Researchers and marketing academics call the tactic "sludge": friction built into a process on purpose, designed to exhaust customers until they give up chasing a refund. AI has not invented sludge, but it has made it cheaper to run at scale.
Ryan Hamilton, a marketing professor and consumer psychology researcher at Emory University, puts it plainly. "Sludge existed before AI," he says. "But AI has just sort of ramped up the dystopian nature of it."
Every call the writer made, to FedEx, to the bike retailer, to his bank, to his credit card company, and to his local police department, routed him into a chatbot: a virtual automated assistant that answers questions and manages queues without human involvement. FedEx's system repeatedly ignored requests to transfer to a person. The police department's chatbot took his information and promised a callback that arrived during a work meeting with no voicemail attached.
FedEx eventually opened a claim and closed it by email, confirming the obvious: the package was missing. The bike company persuaded FedEx to refund the shipping cost, which came to roughly one-tenth of what the writer paid. His bank and credit card company, after their own chatbot mazes, both declined to help because the package went missing on FedEx's side of the transaction. Nearly three months in, he is still out around $1,700.
Companies are making a calculated bet here. Ravi Dhar, a professor at Yale and director of its Center for Customer Insights, describes a sunk-cost dynamic driving AI spending across industries. Leaders who have committed heavily to AI tools face pressure to show results, which makes it hard to slow down even when the customer experience suffers. "If you're a CEO," Dhar says, "you're getting questions from Wall Street like, 'What is your AI strategy, and is it showing any return on investment?'"
Hamilton warns about a longer-term flattening effect. If every company deploys the same AI call centre, service stops being a way to stand out. "Everyone is going to have the same AI call center, no matter what industry you're in," he says. Our 14 July report on enterprise AI agents found that two-thirds of firms are already racing toward zero human oversight of AI decisions, which suggests the pressure on human agents is not easing.
The plain judgement here: corporations are not deploying AI customer service because it serves customers better. They are deploying it because Wall Street is asking about ROI, and a chatbot is cheaper than a person. Until that calculus changes, the writer's $1,700 loss is less an edge case than a template.
Should you worry about a lost delivery?
Yes, and preparation matters more than persistence once things go wrong. Screenshot your tracking page the moment delivery is confirmed, capturing any signature name or initials. File any police report in writing rather than by phone alone, so a paper trail exists. When a chatbot deflects a request for a human, try repeating "agent" or "representative" each time it loops back.
