Citi to IBM's Watson: Find something to do

Home - Viewing one post

March 6, 2012

Jeopardy book

IBM and Citibank announced yesterday that Watson will be working as a banker. Even after reading the IBM news release or the Bloomberg story, you'll only have the haziest idea of what work the computer will be doing. That, I would imagine, is because nobody really knows. It's like a football team that hires a fleet-footed seven-foot-tall 350-pound athlete from Ukraine or Zimbabwe. He might not know how to play football yet, but he has talent. And by experimenting, there's a good chance they'll find a position for him.

One thing to keep in mind is that a Watson-branded technology is likely to be very different from the machine that triumphed in Jeopardy a year ago. The Jeopardy system was a rich combination of technologies, including natural language and datamining, along with multiple flavors of machine-learning, and advanced analytics. When developing AI systems for work in hospitals or banks, IBM can mix and match pieces of Watson with other available technologies, including the analytic systems already in place in those industries.

I wouldn't be surprised to see a "Watson" system that greets visitors to Citibank and lines them up with the right department or perhaps reviews people's investments, and signals areas of high risk. This wouldn't match the sophistication of the Jeopardy system. But it would extend the Watson brand, which is clearly one of IBM's goals. At the same time, far from the public view, the research side of Watson could be "reading" vast libraries of financial documents, correlating its "learnings" with market data, and guiding investment strategy. For Watson to pay dividends in such higher-level jobs will require training, which is what Citibank and IBM will no doubt be providing.

Meanwhile, another version of Watson could be reviewing bank documents and communications, searching for indications of insider trading, and policing corporate governance.

The point is that a system with a combination of natural language, datamining and analytics will find plenty of work in today's data economy. Some of the early efforts will be targeted more toward branding. But in time, Watson-like programs will redefine the work many of us do. And IBM won't be the only company creating such technology.

Update: Here's an IBM blog post that provides more scenarios of what a banking Watson might be up to.