Walmart's AI Service Loop Cost It a New York Times Writer

A Wirecutter writer spent more than an hour trying to reach a human at Walmart customer service, failed, and then deleted her Walmart account — and told the story in the pages of the nation's largest newspaper. The incident, described by Ivy Liscomb in a piece for the New York Times-owned review site, is a working case study in what happens when an automated support system has no reliable path to escalation.

Liscomb's problem began simply. Walmart no longer carried a product she wanted, so she tried to cancel her delivery subscription online. The interface offered no button to do so. She called customer support instead, and the automation would not let her out.

What actually happened on the call?

"A bright, automated voice insisted that I could cancel my subscription online, even after I explained that I couldn't," Liscomb wrote. "At some point, Walmartbot appeared to accept that I was 'experiencing a glitch.' And then it dropped the bomb: I could reload the page and try again, or delete my Walmart account."

She then tried the familiar escape maneuvers: saying the right word or phrase, pressing the right buttons, hanging up and calling back. None worked. "Finally, after spending over an hour attempting to get help from a human, I deleted my account," she wrote. "In an age of countless digital storefronts vying for my cart, it seemed unbelievable that such a successful brand would allow faulty AI to cost it a customer."

Two interface failures compounded each other. The bot rebuffed every attempt to reach a person, and it also forgot her issue after every few questions — no persistent context, no transfer with history attached. Walmart, the second-biggest company on earth by revenue, recently edged out by Amazon on the Fortune Global 500, lost a customer over a missing cancel button and a chatbot with no exit route.

What do the numbers say about escalation?

The survey data around this failure is unambiguous. Gartner reported that 87% of consumers say it is essential for companies to provide an option to reach a human agent when using generative AI for customer service. This week, Consumer Cellular released a survey of consumers aged 50 and older: more than 80% said they would consider leaving a company if its automated system kept them in a loop.

The picture is not uniformly hostile to AI. In the same Gartner survey, half of consumers said their interactions are easier with generative AI than with humans. The technology is not the problem; the missing escape hatch is.

Why do bots refuse to give up?

Customer experience tools are built to resolve as many queries as possible without human help, and that design usually pays off: they move customers between steps with very little wait. The same persistence becomes a defect when no solution exists. The bot keeps searching for an answer it cannot find — a "never give up" attitude taken to an extreme, as Nextiva CEO Tomas Gorny, who co-authored the analysis of the incident, frames it.

The fix is a multistage escalation protocol:

• When a customer expresses any frustration, a human agent is alerted and oversees the interaction, guiding the AI toward the right solution.
• If that fails, the call transfers instantly to a human agent — with full context provided, so the customer never repeats themselves.

Neither stage appears to have existed in Liscomb's calls.

Who can act on this?

Small and medium-size businesses can treat this gap as an opening. Companies that pair effective AI tools with seamless access to real people are positioned to win customers the giants are losing — customers who, as the survey figures show, are ready and willing to leave a brand that does not keep earning their business. As automated support becomes standard, the businesses that master escalation, not the ones that deploy the most automation, will hold the advantage.