DISCUSSION

Bad AI Bots Anger Customers, Yet Brands Must Stand Behind Chatbot Promises

Written by Nicholas Morine

A Nov. 4 Forethought AI Agents by Zendesk study titled “State of AI in CX 2026: Holiday Report” surveyed more than 1,000 U.S. adults on questions related to shopping in the age of artificial intelligence, and the results delivered mixed narratives.

First up, nearly two-thirds of shoppers polled (62%) said they would consider allowing an AI model to make some direct purchases – but there were some major caveats: the most obvious being that only a slim minority (18%) of respondents would let AI buy something worth more than $500.

One major throughline coming to the fore inside of the survey results was that shoppers are quite obviously fed up with inept chatbots or AI customer service agents that aren’t able to properly address their needs – and that shoppers want a human customer service touchpoint when desired.

Some data points backing this thrust included:

  • Cart abandonment is real: Nearly half of shoppers (41%) said they had abandoned an e-commerce cart due to a chatbot failing to answer a simple question related to inventory or shipping. And almost three-quarters (72%) admitted to having experienced frustration resulting from interactions with a chatbot that failed to comprehend their issue.

  • A repeating trend of needing a human appears: While no longer the top annoyance, the No. 1 frustration in the 2024 and 2025 versions of this study was “challenges reaching a human when needed.”

  • Patience is short during the holiday sales spree: About 44% of shopper said they’d wait less than half an hour for support during the peak holiday shopping season before looking elsewhere or canceling their order entirely.

Another major angle explored in the report data was one of accountability, with more than three-quarters (82%) of those polled saying they expect brands to provide a credit or refund if their AI models make promises they fail to keep. And about 15% said that buying the wrong item based on a poor AI recommendation “would ruin their holiday shopping experience more than anything else.”

“A promise is a promise, whether it was made by a person or by AI,” said Erica Clayton of Forethought AI Agents by Zendesk.

“The goal shouldn’t be for AI to never make mistakes. That’s not realistic. AI should only commit to what a business can actually deliver. That means grounding answers in approved policies, checking live order data before quoting delivery dates and handing off to a human with full context when an issue falls outside the AI’s scope,” Clayton added.

It’s not all bad news for proponents of AI in the e-commerce world: On average, shoppers expect AI to save them $192 this holiday season; more than half (57%) place trust in AI to conduct accurate price tracking; roughly 42% trust AI to aid in a search for coupon codes; and 53% of those polled expect AI to cut at least an hour off their holiday shopping duties.

Questions for further discussion:

  • Should a brand or retailer be responsible for honoring promises made by its AI in terms of offering refunds or a credit? Why or why not?

  • If only about one-fifth of shoppers are willing to let AI make purchases of $500 or more on their behalf, what do you believe it would take to convince skeptics to sign on? Is this even possible in the near future, in your opinion?

  • How vital is human oversight, or having a human customer service agent at the ready, for brands and retailers deploying AI chatbots in customer-facing roles? Do you believe AI chatbots are a net positive or negative for retailers and brands in general?

Discussion Thread0

The RetailWire Daily

One morning briefing: the retail news, expert perspective, and business context to bring into your next meeting.

Want to know more? Learn more