dcsimg
 

IBM, NVIDIA, Hyundai, Geico Bring Deep Learning AI to Call Centers

Friday Mar 15th 2019 by Rob Enderle

Call center AI implementations are creating huge improvement rates in multiple metrics, and users are impressed.

Next week, I’ll be moderating a panel at NVIDIA’s GPU Technology Conference (GTC) on artificial intelligence (AI) and how it will massively change call centers. Call Centers are one of those never-ending nightmares that we don’t talk much about, even though they tend to do horrid things to our firms’ reputations and they are a nightmare to keep staffed because working in a call center has to be one of the worst jobs in the world.

Last month, IBM CEO Ginni Rometty was on stage during her Think Conference keynote with auto insurance company Geico, and they spoke to how they used Watson (arguably the most advanced deep learning AI currently in production at scale) to fix this exact problem. To help me prepare for my panel next week, I thought it would be good to revisit this problem and the use of AI in the call center and how it changes that call center from a secret liability to an undeniable asset.

Why Call Centers Are a Pain

If you don’t get that call centers are a pain all you need to do is talk to someone who has had to keep one staffed. Employee problems and turnover are at nearly unbelievable rates and can exceed 200 percent per year in some locations (that means that every six months, you basically have a brand-new call center and always have way too many untrained call center employees). Particularly in large call centers, you have lots of interpersonal, health and behavioral problems. IBM had Hyundai on stage, and they said their call center turnover was 40 percent before Watson (after implementing a Watson solution, it dropped to 10 percent, which is still higher than I’d like). (Hyundai also implemented IBM’s Hyperledger Blockchain solution, which did wonders for its transactional back end, providing vastly higher security and transaction integrity.)

It is interesting to note that this wasn’t Hyundai cars, but its credit card division, and the Hyundai executive was somewhat of an internet star. His important advice was that no one can give you a generic AI (at least not yet) and that if you approach this effort tentatively, and/or in parts, you’ll fail. Better to fully embrace the move and put the focus on getting it done, not arguing whether it should be done.

Offshoring the effort to India or another country where labor is cheap just introduced accents and an even greater disconnection between the firm and potential customers, driving down customer satisfaction and driving up customer churn.

Geico and IBM

Geico is big and selling car insurance (any kind of insurance) is a real pain because there are a lot of shoppers, most of whom aren’t experienced, and a lot of distrust between these shoppers and insurance companies. Insurance is gambling backward. When you gamble, you typically bet that you will win and the house bets you will lose, and the odds significantly favor the house, but you still hope you’ll win. With insurance, you bet you are going to lose (be damaged), the house (the insurance company) bets you are going to win, and you hope you’ll lose because no one wants the damage. The best example of the latter is life insurance because you can’t spend the money if you are dead. But Geico’s VP on stage said that by putting Watson in its call center for a renters’ insurance offering, the company improved close rates by 40 percent, which is huge.

In this implementation, callers are talking to Watson and, apparently, the experience is so real that one of them (the calls are taped for training and litigation protection) said something to the effect that “us late night workers need to work together,” evidently convinced that not only was Watson a real person, he’d like that virtual person as a friend. Geico did set a very low bar for this effort and that bar was “do no harm.” But the company massively exceeded that bar and is now a huge Watson Call Center advocate, based on what was said on stage.

One interesting comment from IBM’s Rometty was that the best approach they’ve discovered for doing this across the many customers that have done it is to redesign the call center around Watson, not layer it on. Doing it this way, they saw the greatest positive impact on closed deals.

Another interesting comment came from the Geico VP on Watson in particular and digital transformation in general. He indicated the biggest problem was that there are a lot of folks who hate change. To get them to change, you need lots of data and some strong, believable third-party advocates who back the data up. I would add that you need an executive with actual power who is willing to put their name behind the project.

Wrapping Up: Improvement Rates Suggest We’re All Going to Love AI

Call centers are a nightmare, people hate the job, customers aren’t fans either, and an out of control call center does massive damage to your brand and revenue potential. This appears to be an ideal place to put AI because it never tires, it learns on the job at machine speeds (if it is a deep learning system like IBM Watson), and it shows a path to fixing this AI problem. That Geico close rate improvement of 40 percent was a massive improvement, as was the drop in employee churn for Hyundai from 40 percent to 10 percent. And realize that Watson is still relatively young. Imagine what you’ll be able to do 10 years from now. We all may not only be happy to give Watson the job, the customers that use the resulting service may love it even more.

If you are at NVIDIA’s GTC, my panel on this is at 11 AM on March 19. Drop by and say hello.

 

Home
Mobile Site | Full Site