From hotels to retailers, the vast majority of businesses are deploying large language models, with primary use cases covering search and customer support. However, these models are not flawless.

Consumers are frustrated with AI chatbots that cannot answer their inquiries, and patience is quickly running out.

According to research released by Parloa on Monday, some consumers will simply give up. Three out of five consumers abandon automated support after repeating their question once.

These friction points can be costly. Here are three key figures on how AI is letting customers down:

50%

According to a survey of 1,000 U.S. consumers released by Kim.cc last week, when AI support fails to answer a customer's inquiry, 50% of consumers blame corporate leadership.

When consumers face AI chatbots or tools that cannot resolve their issues, they are not forgiving.

"The most important takeaway for leaders is that customers do not view poor AI support as a technical issue, but as a company issue," Sachin Jaiswal, co-founder and CEO of Kim.cc, said in an email to CX Dive.

"If a customer gets a wrong answer, gets stuck, or feels the company is hiding behind a bot to save money, the loss goes far beyond a bad ticket," Jaiswal said. "It can lead to loss of trust, damage to brand reputation, and ultimately losing that customer."

The frequent friction caused by AI is likely the reason why more than a third of respondents immediately feel frustrated when customer support is driven by AI, according to the Kim.cc survey.

91%

According to an analysis conducted this spring by AI startup Oumi for The New York Times, Google's AI model has an accuracy rate of 91%.

Google processes more than 5 trillion searches annually, which means that even with a 91% accuracy rate, it still provides billions of incorrect answers each year. One of those incorrect answers recently made headlines. Earlier this month, a regional court in Germany issued a temporary injunction against Google regarding false information generated by its AI Overviews feature. Two companies filed lawsuits because certain searches caused Google AI Overviews to incorrectly associate them with fraud and unfair business practices. The court ruled that the false and defamatory statements generated by the AI Overviews feature constitute the company's own content, and therefore the company is responsible for it. Although Google is contesting the ruling, the tech giant is currently responsible for 80% of the legal costs. If the ruling stands, it will open the door to more lawsuits and defamation cases against tech companies relying on generative AI.

$5,000

A car dealership paid a $5,000 price for its out-of-control AI chatbot incorrectly overvaluing a car.

According to a CBC News report last month, when a Toronto man decided to sell his car, he submitted an inquiry to the BMW dealership where he had purchased it. He received a text message from "Quinn" offering a buyback price of 27,162.79 Canadian dollars (approximately $19,119 USD). It turned out Quinn was an AI chatbot, and the offer was a mistake. The dealership withdrew the offer until CBC News contacted BMW Toronto for comment. The dealership then reinstated the original offer, paying the man about 7,000 Canadian dollars (approximately $5,000 USD) more than the dealership actually believed the car was worth.