A Lost Ebike, FedEx, and the Rise of Chatbot Sludge

A $2,000 Ebike Vanishes

The trouble started with a splurge: a pair of ebikes, nearly $2,000 per bike, ordered online for tackling the hills of Atlanta. One arrived without issue. The other, ordered from a different retailer, was delayed repeatedly.

Then, one Wednesday evening, a text from FedEx said the second bike had been delivered and signed for. The author was in the kitchen at the time, very much bike-free. The tracking details showed a mysterious signee with the initials “M.M.”—no match for anyone in the household or building.

The package was gone. What followed wasn’t just a dispute over a shipment, but a months-long trip through automated customer service systems across FedEx, the bike company, the bank, the credit card issuer, and even a local police department.

When Customer Service Becomes “Sludge”

Companies have been aggressively rolling out AI in their customer support operations, often while shrinking or reshuffling human teams. In a survey of customer service leaders published in April, 31 percent said they have already reduced or are planning to reduce headcount due to AI adoption. Most say they’re moving agents into new roles or piling on new tasks instead of outright layoffs.

Some executives are far blunter. Verizon CEO Dan Schulman recently told Bloomberg that AI will likely replace a “large percentage” of the company’s customer service work, calling it one of the most exposed sectors.

Recommended reading

Mira Murati’s lab unveils Inkling, a 975B open model

For consumers, this has meant fewer humans and sharper versions of familiar frustrations: long waits, canned replies, dead ends. Researchers say this often isn’t accidental. It can be a deliberate tactic known as “sludge”—designing friction into processes to discourage people from pursuing resolutions.

Ryan Hamilton, a marketing professor and consumer psychology researcher at Emory University, argues that AI has simply supercharged an old tactic.

“Sludge existed before AI,” Hamilton says. “But AI, like with everything else, has just sort of ramped up the dystopian nature of it.”

A report published in May covering consumers in the US, UK, and Canada found 59 percent were frustrated with AI customer service agents, and 85 percent said they’d rather talk to a real person.

Chatbots at Every Turn

In the missing-bike case, nearly every call hit a chatbot wall. FedEx’s AI agents repeatedly ignored requests to reach a human. Even filing a police report meant interacting with automation.

Calling the local police department to report missing property led to a prompt to leave information with a chatbot and wait for an officer to respond. Two attempts followed: one report filed over the phone, another through the department’s website. The first never generated a response. The second produced a call back during a work meeting; when that call was missed, there was no voicemail, and trying to call back dropped the author into the same chatbot-driven non-emergency line.

FedEx eventually opened a claim and, via automated email, confirmed the bike was missing—but directed the customer to seek restitution from the shipper. The bike company did pick up the phone with a human, but could only get FedEx to refund the shipping cost—about one-tenth of the total purchase.

Attempts to dispute the charge with the bank and credit card company also funneled through “long, chatbot-filled rabbit holes” that ended with a human agent saying they couldn’t help because the loss occurred on FedEx’s watch.

Agentic Hype, Messy Reality

The technology at the core of these systems is still early. Agentic approaches have attracted heavy hype, but even Meta’s Toolformer—touted for its ability to call external tools when needed—is only three years old. A more capable system might have at least routed the case to the right person or used FedEx’s extensive shipping data to locate the bike.

What’s unclear is whether companies will actually deploy more advanced tools in ways that benefit customers, rather than just cut costs.

Hamilton says some firms roll out agentic customer service bots without fully grasping the hit to customer experience.

“Sometimes they’re aware,” he says. “They’re willing to accept that trade-off.”

He warns that some sectors can’t afford degraded support, and that as AI-led interaction becomes standard, businesses will need other ways to stand out—possibly by offering better service. Otherwise, he says, there’s a real risk of:

“smoothing out the service dimension, where everyone is going to have the same AI call center, no matter what industry you’re in.”

Sunk Costs and Investor Pressure

Ravi Dhar, a professor at Yale and director of the school’s Center for Customer Insights, sees a sunk-cost dynamic pushing AI deployments across industries. Global spending on AI tools is ramping sharply this year, potentially locking leaders into strategies even when the returns are murky.

“If you’re a CEO,” Dhar says, “you’re getting questions from all of the investors, from Wall Street, like, 'Hey, what is your AI strategy, first of all, and is it showing any return on investment? You’re spending all this money.'”

Hamilton believes many leaders are operating on “optimism,” betting that going all-in on AI chatbots will eventually work out, even if it hurts in the short term.

“They kind of assume that AI will catch up, or it won’t be that bad,” he said. “And it can, in some circumstances, be quite bad.”

Three Months Later, Still No Bike

Nearly three months after the supposed delivery, the bike remains missing. FedEx never located the package, and the only reimbursement was the shipping refund—leaving the author about $1,700 out of pocket.

When contacted, the Atlanta Police Department said they had in fact dispatched an officer the day the earlier call was missed and promised to send one again. They clarified that an operator, not an officer, had placed the original call—and that returning the phone number routes people back to the same chatbot-run non-emergency line.

FedEx, in a statement, defended its use of automation:

“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 amplify 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.”

The missing ebike is still unresolved. The chatbots are still answering the phones.