Yesterday we had an analyst briefing where we went deeper in what we currently do with clients and some of the fundamental innovations behind Watson. The marketing may sometimes sound to glorious to be truthful and creates expectations that are too high – like – oh my god, if Watson can do anything and answer every question, no expert is needed anymore. Well that’s not the case…
But it can enable companies do much more work to get better and more charmingly engaged with clients – than with today’s kind of online self-help or call center service. To do so, Watson needs a couple of capabilties. Some of these were built in a scalable and componentized way for the Jeopardy! game, some need to be adapted or created for each new domain where Watson goes to work. To give an example, Watson needs to be trained on the specific domain of knowledge (e.g. medical expertise on cancer diagnosis and treatment or customer service in telco or banking) plus on the company specific knowledge and language. That’s an area where Watson indeed goes to work now, see e.g.: Watson at Your Service: IBM Unveils the IBM Watson Engagement Advisor and the video Announcing the IBM Watson Engagement Advisor.
I’ve just come across an amazing example of a superior customer centric service – delivered in a 21st century social business fashion: the McDonalds Canada “Our Food, Your Questions” program launched in June 2012. See this great summary article: “Marketing, With A Side Of Truth: The Secret McDonald’s Recipe For Canadian Success”.
I would not expect Watson to answer all of these questions – a lot of those are requiring research and knowledge acquisition in the first place – which is what makes this really authentic. But given the enormous amount of questions asked and answered before, this would make a great source for Watson to reply on questions that are similar to some asked before – based on the understanding of the meaning of the question, not just the keywords. Just think of the shortcomings of searching for an answer that’s certainly be answered numerous times before. Answers spread in many places… and maybe not retrievable by a simple keyword search. This is the kind of cognitive computing that Watson can do… given some preparation and training. And more work and effort could be done by the company (McDonalds in this purely hypothetical and illustrative example) on researching for answers where Watson doesn’t find a clue in the existing corpus of knowledge.
“Science determines the limits of the possible. Engineering lets us reach them”, Clifton Leaf.