Thursday, July 25, 2013

Can Watson actually replace customer service agents?

By: Alon Cohen EVP/CTO

I was asked to opine about the above question "Can Watson (the Supper Computer from Jeopardy) actually replace customer service agents?".

I have no doubt that computers will replace humans in many jobs in the time to come. I have seen the airline industry move from 5 person crew: a Radio Operator, Navigator, Flight engineer, and two Pilots, to a 4 person crew as radio became simpler to operate, then to a 3 person crew as GPS and INS (Inertial Navigation System) became simpler, and currently only a two pilots crew when computers replaced the Flight Engineer.

Interestingly a pilot I flew once with, inside the cockpit of a Boing 767, told me that soon there will be only one pilot and a dog. When I asked the obvious “why the dog?” the pilot said, “well, in order to bite the pilot if he tries to touch anything”.

However, I think the future is not that grime for the Flight Engineers and humans in general specifically if you notice the fact that Scotty the flight engineer is still there in Star Trek.

Anyway if you like to read more about Watson, Speech Recognition and my take on using Watson for customer support check out my article in SpeechTechMag:

Can Watson actually replace customer service agents?

Watson, IBM’s cognitive computing system capable of answering questions posed in natural language, has been helping professionals in the clinical medicine and financial services fields diagnose ailments and recommend products that save companies time, money and resources. Now, Big Blue would have Watson put on a new suit: customer service agent.

As an executive who is rooted deeply in providing high-quality, consistent customer service and as a technologist who believes we humans someday will indeed interact with computers via the spoken word, I have to challenge Watson’s newest use case.

Based on what I read, IBM chose to incorporate existing speech-recognition technology (like the one used for Apple’s Siri) as a front end to Watson. This will require near-flawless speech recognition, a technological challenge that has been very elusive for many years now. 

Apple with Siri and Google, at least publicly so far, haven’t been able to provide a solid solution. 

The fact is,  speech recognition is still one of the most challenging technological hurdles to overcome, but there are a few possible workarounds. One is to use text input, which we know is kind of slow and annoying for users.

Watson, however, could someday become the ultimate customer service agent with the help of humans. First, by using people -- biological speech recognizers, if you like -- as listeners, who would type in customers’ initial questions and comments to Watson. These people would not talk. They would merely listen to the callers and type what they hear to Watson so it can talk back to the customers using text-to-speech technology, thus providing the illusion of natural-speech interaction with Watson. The listeners could be located anywhere in the world and reliably translate different languages.

Next, in addition to listeners, a company should have a person or a small group of experts training Watson to prepare the system for acting as a huge number of support agents handling customer issues at once. 

Training would involve product and technical information, even to the point of teaching new words or sounds related to the company’s products or services. 

Such training could potentially solve the wait-time problem and variations in expertise from agent to agent that every customer-support organization suffers from. 

Of course, using behind-the-scenes human listeners to complement Watson and investing in customized training to fine-tune its applicability to a specific company would mean expense in addition to the acquisition of the system.  On the other hand,  Watson will never ask for a sick day or have to take a vacation, an “always-on” capability that would go a long way toward achieving consistent and high-quality customer support.

Since human listeners might be needed in order for Watson to interact accurately with customers, I do not see this technology lowering the cost of customer service just yet. However, the quality of service might go up dramatically, increasing brand loyalty.

Customer service is truly an art form and, like many executives across the globe, I welcome any new solutions that will help make it better. I plan to keep my eye on Watson. But whenever a fully automated or human-assisted speech recognition solution is someday achieved, the track record of reality trailing rhetoric leads me to think it will be several more years before a Watson-based CMR solution replaces human agents in headphones.

Meantime, I really hope that the Google Glass technology that relies on speech recognition will set new and higher thresholds for speech recognition that will finally make Star Trek–level, speech-based person-to-machine interaction a commercial reality.