IBM’s Watson Is Bringing “Cognitive Computing” to Customer Service

It will be powering smartphone apps by the end of the year too.
Clockready via Wikimedia

IBM’s Watson computing platform made a name for itself on Jeopardy, but its incremental roll-out into the real world has been no less impressive. It has worked in finance at Citi helping to assess risk and at Memorial Sloan-Kettering Cancer Center sifting through medical cases and data to help oncologists make the right diagnoses. Now the supercomputer is rolling out to the masses as a computerized customer service agent designed specifically to help customers connect with the information they want at a variety of firms in a variety of businesses, including Australian finance house ANZ, at media ratings maker Nielsen, and at Royal Bank of Canada.

This new offering of Watson, known as IBM Watson Engagement Advisor, will give customer service transactions a layer of cognitive computing help, leveraging Watson’s unique skills to make those transaction go more smoothly. Those unique skills include Watson’s ability to understand context and to learn as it goes, giving it far deeper insight and the ability to return far more meaningful answers than a simple Web search can.

The reason Watson can do this is because it is more or less designed like a human brain. That is, where a Web search is great for finding certain blocks of text out there on the Internet or sifting through structured data sets, Watson works well with unstructured data, or data that hasn’t been organized into a database. Web search can find you a snippet of text within a sea of words as directed by the terms of your search. Watson can understand both the nuance of your question and the context of the text surrounding your potential answer, wherever it lives out there on the Web. As such, Watson’s answers tend to be more meaningful.

On top of that, Watson can learn as it goes. So imagine Watson working alongside your financial planner. Your planner asks you questions, you answer, Watson remembers your answers. Watson is also sifting through an Internet’s worth of financial data as well as your specific financial data, looking at your transactions and the economy in your city and state and the global economy and several anecdotal, unstructured stories and opinion on the municipal bond market, in which you have invested a meaningful portion of your savings. You can ask your advisor a specific question about your portfolio and its prospects for the near term, and he or she might have to get back to you with the answer. Watson can put a good answer in front of the both of you instantly.

And it doesn’t stop there. Beyond being a client-relations assistant, IBM is also bringing Watson directly to the masses via an “Ask Watson” feature that offers help to customers via various channels, including email, text, and chat. Considering nearly half of the 270 billion customer-service-related calls go unresolved every year, Watson’s cognitive ability could go a long way toward alleviating load stress on call centers and in resolving customer service issues (imagine calling the help desk only to find you can’t get a live person on the phone and you still get to walk away satisfied).

IBM Watson Engagement Advisor will roll out with a handful of companies (including those listed above over the next few months, during which it will be evaluated for a variety of tasks (ANZ will use it to evaluate insurance portfolios to help customers determine where they are over-insured and where they are exposed, while Nielsen will roll it into the software tools that help its clients in media planning figure out where to buy their ad space) before being released into the larger ecosystem.

Keep an eye on it, as it will likely grow into something larger, and quickly (IBM says the first Watson-powered consumer apps will emerge later this year). Imagine a voice-activated Watson assistant that understands your linguistic nuance and learns from your past queries what to expect in the future. In other words, imagine a Siri that really works.