IBM's Watson gets a job in customer service
By David Szondy
May 23, 2013
IBM’s Watson supercomputer has been riding high for the past couple of years. It won a game of Jeopardy, went to university and did a stint at a cancer lab. But now it’s taking what might seem like a step down with a job in customer service. According to IBM, the current avalanche of information is provoking an oncoming crisis in customer service and the company sees Watson’s advanced learning and data crunching abilities as a solution.
If you've ever been stuck on the phone for an hour with a help desk person who seems more lost than you are, then you know how important customer service is. One bad experience can lose a customer for good, but effective customer service isn't cheap and companies often end up relying on workers at overseas call centers working from scripts that they don’t really understand and follow by rote.
This is now so commonplace that it’s become a comic cliché, but for customers it's far from funny. Today they have higher expectations and demand more personalized service with more personal interactions and self-service options. Worse, the “Millennial” generation will soon become a major customer group with a taste for mobile devices and a desire for fast, interactive services. That means even higher expectations and more willingness to drop a company that can’t satisfy them.
The other side of this equation is the growing information glut. IBM says that the amount of data available today is growing so fast that 90 percent of the world's data was created in the last two years. Additionally, 90 percent of all the data is unstructured, putting it beyond the reach of traditional computing systems that prefer data in neat databases. That can make finding the relevant information a time consuming process that fails to an alarming degree. According to IBM, 270 billion customer service calls are handled every year, yet only about half are resolved, resulting in customer service costs tripling. The company also claims that 61 percent of these unresolved calls could have been handled with better access to information.
This is where IBM Watson comes in. The company sees the supercomputer as part of a new era of cognitive computing systems and its involvement on Jeopardy two years ago was a test of the machine's ability to emulate human cognition. For the game, it had to react quickly, processing large amounts of data while making complex and subtle logical connections. It also had to handle double meanings of words, puns, rhymes, and inferred hints, all tasks that are second nature to humans but have long stumped computer systems.
IBM Watson works by moving beyond keyword searches and structured databases by employing natural language processing. In the case of the quiz show, it can deal with the subtleties and ambiguities of speech. It can also generate and evaluate hypotheses by weighing possible responses based on only relevant evidence. And finally, its evidence-based learning lets it improve its performance based on outcomes, so that it “gets smarter” with each iteration and interaction.
The clever bit is that what works for answering questions about American presidents and Hollywood celebrities also works in customer service. And since its Jeopardy appearance, Watson’s technology has improved to the point where it enjoys a 240 percent improvement in system performance while the physical requirements have been reduced by 75 percent, meaning it is now powered by a single Power 750 server running Linux. The upshot is that a supercomputer that was once the size of a master bedroom is now as big as four pizza boxes.
The version of the computer intended for customer service, marketing and sales is called the IBM Watson Engagement Advisor. It can be used either by customer service agents or through a customer’s mobile device via cloud-delivered services and online chat sessions. IBM says that the system learns, adapts and understands a company's data quickly, while increasing its knowledge and capability over time. This allows it to address customers' questions, offer feedback to guide purchases and decisions, and troubleshoot problems.
"IBM is taking Watson to the masses in ways that positively impact people's lives, from helping doctors improve patient care to helping businesses put consumers first, in an increasingly mobile world," said Manoj Saxena, General Manager, IBM Watson Solutions. "Customer engagement is a natural fit for Watson, which can instantly create a strong bond between who customers are as individuals, and what types of information will help them reach their goals."
IBM sees the financial and health fields as the initial customers for IBM Watson Engagement Advisor, with the first adopters being ANZ bank, Malaysian mobile telecommunications company Celcom, IHS Inc., Nielsen and Royal Bank of Canada.
The video below outlines the features of the IBM Watson Engagement Advisor.