How to handle angry customers, AI version

The research project will use machine learning to mimic furious customers and run through millions of scenarios to figure out how customers react in different situations.

Companies often frustrate their customers -- or make them downright irate -- because of product problems, shipping snafus or rude people on customer service lines.

The problem for companies, though, is to figure out exactly why their customers are angry before they ditch them for a competitor.

To help enterprises discover the root causes of customers' anger, researchers at a New Zealand-based company are using artificial intelligence (AI) and machine learning to mimic furious callers and run through millions of scenarios to figure out how customers react in different situations.

"These are often the seeds of opportunities where a business can differentiate itself in the market by consistently delighting customers in ways not easily visible to competitors," said Frank van der Velden, CEO of Touchpoint Group, a software company focused on the New Zealand and Australian markets. "One of our key objectives is to automatically detect these situations, and to allow both risks and customer opportunities to be quickly acted on."

Touchpoint is putting $500,000 into the project, which has been dubbed Radiant after a predictive supercomputer in sci-fi author Isaac Asimov's Foundation series of novels.

The research project will run what-if scenarios to see if a given situation is likely to enrage or please a customer, according to van der Velden.

The AI program will go to work over the next six months, simulating hundreds of millions of angry customer interactions.

It also will use what Touchpoint calls a "massive" data set to come up with thousands of different customer experience variables to help Radiant learn to predict customer reactions to various situations.

According to Touchpoint, it will test millions of these scenarios each week.

"Figuring out what makes customers angry and why is a pretty good use for AI," said Dan Olds, an analyst with The Gabriel Consulting Group. "Assuming the models used by the AI are accurate, the results should provide valuable insight for businesses into not only how to avoid making their customers angry, but how to best fix it when they do become mad at the company."

He added that it's not that enterprises don't know what upsets customers The problem is more that they don't know what angers customers the most and how to best fix the mess.

"Do customers become more angry over products delivered late or products that don't work as advertised? Acquiring a customer is expensive, and businesses need to do everything they can to keep a customer buying from them for a long time," said Olds. "I think the combination of enterprise IT and machine learning is just barely entering its infancy. In the future, I expect that most enterprises of any size will have some sort of functions that make use of these deep learning concepts."

Jeff Kagan, an independent industry analyst, said he wishes companies would focus less on tech like AI and focus more on simply talking to their users or customers.

"There is a problem with enterprises not having the courage to have a conversation with their customers," he noted. "On one hand, I am very happy that companies are interested in learning about how they tick their customers off. On the other hand, why the hell don't they just ask their customers? Using AI to measure what is angering customers is like using a hacksaw instead of scissors for a haircut."

However, Rob Enderle, an analyst with the Enderle Group, said companies have been trying to figure out what is putting off their customers for a long time now. In previous attempts at this type of research, they often go in with preconceived notions and taint their own results.

Using artificial intelligence could fix that problem.

"There has been a ton of research done on this and the vast majority has been a waste of money," said Enderle. "A properly configured learning system that can emulate actual interactions should be able to more accurately determine problem areas and then be used as a quality testing step to assure future systems don't have these problems."

Join the TechWorld newsletter!

Error: Please check your email address.

Tags popular scienceemerging technology

More about

Show Comments
[]