Can Car Dealers Trust AI?

Artificial intelligence is no joke. But it needs to work on its sense of humor, at least when it comes to responding to customer reviews of car dealerships.

That’s the word from Matt Murray (pictured, below, left), CEO of Widewail, and Max Muncey, corporate communications director for the Michigan-based 35-dealership LaFontaine Auto Group.

Matt Murray

“We want to fully understand the customer experiences, and reviews help us with that,” Muncey says.

Reviews are so crucial to the LaFontaine organization that CEO Ryan LaFontaine gets a daily summary of them.     

Widewail’s business includes soliciting reviews for dealer clients, monitoring them and responding to them whether they are on dealer websites or social media sites such as Google.

Murray believes in the future of AI, but he’s a realist about its current state.

“Because of the impact of artificial intelligence, we are investing a ton of our time and resources on it,” he tells Wards. “We have a lot of faith in AI as a quantitative tool. But right now, it is in no way a standalone tool.”

So, although AI generates suggested responses to customer reviews, humans at Widewail look at and often revise them before they are posted.

 “I just can’t imagine something so important to dealers as customer experience being completely managed by an unmonitored machine,” Murray says.

To its credit, AI demonstrates an impressive ability to handle large amounts of data.

“We’ve used it to process 800,000 Google reviews for dealers,” Murray says. “All of this unstructured data is sorted by category, sentiment rating and the density of mentions of certain topics.

Artificial intelligence is unbelievable at undertaking those tasks.”

Similarly, the LaFontaine Group uses AI to analyze trends but not to respond to customer reviews, says Muncey, whose job includes reputation management. That ties into customer reviews.

How AI Falls Short

So why does AI get less than an “A” on first-draft responses?

Murray says: “In managing the customer conversations or reading a review and responding to it, is where we see imperfections.

“We see plenty of ‘hallucinations,’ the use of words or phrases that just don’t make sense in context. Improper grammar also is pretty common, unfortunately.”

Then there are weak attempts at humor. How does AI come up with those? It has to do with human nature.

“Remember, AI gathers its capabilities by understanding human intelligence,” Murray says. “It’s repeating everything we produced over many years to come up with the relationship between words. And people, by nature like to make jokes. What people say affects what AI says.”

Widewail uses a digital language model of what can and can’t be said in automated personalized responses to customer reviews.

“We try to eliminate the use of specific words,” Murray says. “For example, a few months ago, AI kept saying, ‘We’re so thrilled to see your review.’ Or: ‘It’s thrilling to see your review.’

Widewail told the AI system to knock it off. What happened next?

It produced a response that said, “We are thrilled (oops) to see your feedback.”

Murray calls it AI’s idea of a joke.

It’s also why he doesn’t currently see AI at a point where it can go it alone. “It’s a touch irresponsible to allow just the machine to manage these potentially very sensitive customer communications – that also are being seen by potential next customers.”

Consequently, Widewail puts a team member between an AI-generated response and its actual posting to ensure it is accurate, with the right tone and personalization.

Responses to customer reviews “still need a human element,” says Muncey. “We want to make sure the tone and meaning is right.”

LaFontaine assigns “customer-satisfaction champions” at each store whose jobs include handling review responses.   

Customer Reviews Are Everywhere

Customer reviews are ubiquitous today. Few people buy a vehicle at a particular dealership without first checking out reviews from other consumers.

It’s more likely that prospective customers give more credence to reviews than to marketing claims, Murray says.

“Our goal is to push comments to sites that automotive consumers use, such as Google, Cars.com, Facebook, DealerRater,” he says. “About 80% of the time, it’s Google.”

A Widewail team measures sentiments contained in reviews.

Those include topics of discussion. “Are they primarily discussing pricing, or are they upset with F&I, or particularly happy with their salesperson or service adviser?” says Murray, who worked at Cox Automotive’s Dealer.com before he co-founded Widewail in 2018.

No dealer wants a slew of bad reviews, but those can be instructive, Muncey says. “Yes, we love positive feedback, but we can learn from negative feedback, which is not to say we get a lot of it.”

Often, the latter involves a manufacturing issue – such as a microchip shortage – that’s out of a dealer’s control.

“But we want to know if we have a bad apple so we can deal with that,” Muncey says. Conversely, if an employee consistently gets glowing reviews, “we want to know what that person is doing so we can share it with co-workers.”  

Reviews Now Government-Regulated

No company is immune to sour customer reviews once in a while. When a business gets nothing but rave reviews, it raises eyebrows.

“The idea that a store is going to have perfect 5-star reviews every time is a red flag,” Murray says. “That’s not achievable. An occasional negative review, in a lot of ways, lends credibility to the positive reviews.”

It’s not just a credibility issue. It can also become a legal matter.

Murray notes that the Federal Trade Commission will punish enterprises who game the system to show only customer accolades.   

“The FTC really cares about this process,” Murray says. “Its goal is to protect the voice of the customer and the genuine nature of review feedback. Unduly trying to avoid getting negative reviews is expressly prohibited by the FTC.”

He adds: “We’ve seen an increase in legal action pointed at eliminating bad actors and processes that suppress negative reviews. The FTC has been very clear in saying, ‘If you choose to engage in these types of actions, we will come to you.’”

Case in point: The federal regulator last year fined an e-tailer $4.2 million for suppressing negative reviews.       

Muncey says LaFontaine uses its dealership management system to ensure that any review stems from a recorded service or sales transaction.