Can AI Fix What’s Wrong With Customer Service?

Wharton’s Christian Terwiesch talks about the tension between human and AI-powered customer service and why there’s a trade-off between cost and quality.

Everyone has a customer service horror story, and it usually involves a frustrating interaction with an employee or agent.

“Let’s acknowledge it’s pretty damn hard,” Wharton professor Christian Terwiesch said about providing great customer service. “It’s tedious, very labor intensive, and has to be provided on demand and in synchronization with the needs of the customer. It’s nothing you can do on an assembly line.”

Companies are increasingly turning to AI-powered chatbots in the hopes that technology can handle displeased customers better and more cheaply than human agents. But chatbots don’t always get it right, either. Terwiesch thinks that, at least for now, the fix for companies seeking customer service salvation is somewhere in the middle. It takes a combination of tech and people.

“I can give you a glass of champagne, I can give you a back massage, and I can make you the happiest customer. But it’s going to be bloody expensive.”— Christian Terwiesch

Lower Cost vs. Better Quality

“There’s a cost-quality trade-off,” Terwiesch, a professor of operations, information and decisions, said on This Week in Business. (Listen to the podcast.)

Along one axis, there’s productivity that can be measured in dollars per customer-service episode, he said. Along the other axis, there’s a measure of customer delight, willingness to pay, utility, or other related outcomes.

“I can give you an amazing service if money doesn’t matter. I can give you a glass of champagne, I can give you a back massage, and I can make you the happiest customer. But it’s going to be bloody expensive,” he said. “Or you’re going to wait an hour on the call line, and I’m going to give you crappy service with somebody who barely speaks English. There’s this trade-off.”

Companies have been trying to use AI to push out the trade-off curve and provide better service at a lower cost. But solving for lower costs, as opposed to better customer service, is a problem, according to Terwiesch.

He said companies need to determine what they really want to provide, because customer expectations are only getting higher. And the higher the stakes, the more important the human touch, such as in a medical setting or real estate transaction.

“We’re waking up to the reality that ChatGPT and other tools are really good at getting us 80% of the way, but not to 100%,” Terwiesch said. “Most people wave their hands at some point and say, ‘We’re going to need a human in the loop.’ And that is when it gets expensive. That human in the loop is costing a lot of money, and the workflow is even more complicated than before.”

“We’re waking up to the reality that ChatGPT and other tools are really good at getting us 80% of the way, but not to 100%.”— Christian Terwiesch

When Does AI Benefit Customer Service?

The professor said he also sees a lot of companies struggling with how to educate their customers on the self-service benefits of AI so that they can embrace it. Until that transition is complete, keeping humans in the loop may well be worth the cost, he said.

Chatbots and other tools have many customer service benefits, including eliminating language barriers. But Terwiesch said he’s more impressed with what he calls the “eloquence” of AI.

“We’ve seen study after study [that] it’s better at arguing, it’s more empathetic, it’s more lovely to talk with an AI person than a human being,” he said. “That’s scary and tells us a lot about human beings. But it comes up in study after study that AI is really sweet and persuasive.”

When asked what he thinks will become the long-term impact of AI on customer service, Terwiesch doesn’t look too far ahead. For now, a hybrid workflow of human and machine is likely the best scenario for most companies, he said.

This article originally appeared on Knowledge@Wharton.