IBM’s Watson AI is working in a South African call centre

Watson, the artificial intelligence system developed by IBM, is most famous for winning an episode of the hard-as-nails US gameshow Jeopardy. But the firm didn’t spend all that money on a piece of software capable of reacting to natural language queries in order to win back a fraction of the development costs.*

And now it’s here. IBM issued a press release earlier today confirming that Watson is being used by local healthcare company Metropolitan Holdings in order to assist customer service operatives.

You know how frustrating it is when you get through to a call centre, ask what seems to be a run of the mill question and then can’t understand why the agent on the other end of the line doesn’t even know where to look to find the answer? That – as far as we can tell – is how Watson will be used initially here, advising advisors on healthcare for customers.

Since it shamed mere humans into submission on prime time TV, Watson has been put to use in a variety of ways. The system’s ability to parse natural language is pretty much unrivalled, and it’s been tuned towards being used as a decision-making tool. At a recent IBM event in Sandton, we saw a demo of Watson being used to advise doctors on treatments, for example.

The advantage of Watson over other search query tools is that it is specifically designed to work with lots of uncorrelated data sets, and come up with an answer to a natural language question based on as much relevant information as it can find. So in the example above, a doctor goes about his or her daily business checking problems and prescribing medicine, but Watson supplements the doctor’s knowledge and acts as a true outboard brain by throwing up other information the specialist may not have been aware of or has forgotten.

Metropolitan Health’s system here in South Africa, however, will be used primarily by the customer support team initially, assisting with healthcare advice. According to Metropolitan, it currently receives 12 million customer “interactions” a year, and Watson will be trained to pull relevant data during a call.

Of course, that customer support team had better not let it become too efficient, because according to the press release Watson is after their jobs.

As Watson learns, it will be able to anticipate the “questions behind the questions”, and prompt agents to proactively share information. This allows for all relevant information to be provided in one call, which would normally require two or three follow-up calls. This saves time for both customers and Agents. Over time, Watson will be taught to respond where possible, directly to customer enquiries, and the convenience of having access to health and wellness information 24/7 will be made possible.

IBM also made other announcements about where Watson has been deployed globally, including Canada.

* For reference, the prize money for the Jeopardy win was $1m (R10m). It’s almost impossible to put a total cost of development on Watson, since the project draws on previous IBM initiatives like Deep Blue. You need a million dollars worth of hardware to run it, though, and estimates of the project’s development costs are between $900m and $1.8bn. IBM is investing $10m a year in Watson.


About Author


Related News