When the Watson supercomputer conquered the game show Jeopardy!, it was IBM's most notable breakthrough in artificial intelligence since Deep Blue was playing chess in the '90s.
It didn't take long for James Wallis to start thinking about the broader implications of Watson's achievement.
"How can we take those underlying systems-the data management, the analytics capability-and apply it to the real world? Apply it to the benefit of society and apply it to the benefit of business?"
Wallis was understandably enthusiastic as he posed these questions to payments industry executives this spring at the annual NACHA payments conference. He is, after all, vice president of IBM's global payments industry group.
But it's not just IBM folks who see potential value in Watson's mix of data and wits. Executives with Bank of America Merrill Lynch, Capital One and other key players in payments already are discussing how Watson-type technology might revolutionize fraud monitoring, credit evaluation and counterparty settlement, or even sit in for human operators when customers log in for online help chat sessions.
Based on Watson's virtually unlimited ability to gather data and its potential to apply risk intelligence to credit evaluations, "we might be able to-even in this country, where credit is on the deteriorating side right now-approve customers we couldn't before," says Peter Davey, a senior manager for enterprise payments with Capital One, who also spoke at the NACHA conference in Austin, Tex., in April.
It could be a while before Watson-style analysis will be ready for bank sector applications. Other industries already investing in the technology-health care, in particular-are at least two years away from launching commercial applications.
But in his three-day Jeopardy! romp, Watson proved extremely adept at determining the confidence levels for potential answers, an important strength that could apply to weighing the financial sector's persistent risk versus reward quandaries. For example: his level of certainty was so accomplished, Watson assessed complex wagers down to the dollar level (such as the seemingly arbitrary amount of $2,127), instead of rounder and random figures that human contestants use.
In the months before his February tournament showing, Watson was performing at a level where his bets might have driven IBM broke. Brian Dalgetty, director of product management for IBM Software Industry Solutions-the unit responsible for commercializing the Watson technology-notes that in these practice sessions Watson was right less than 30 percent of the time (Jeopardy! grand champions are about 97 percent accurate). But the algorithms that gave Watson the ability to intuitively weigh all answers from possible solutions also clued the machine into how it was making errors, providing more confidence to responses and more context to the flow of actual human language.
"That's the fascinating thing, how quickly Watson can learn," Dalgetty says.
Cindy Murray, head of global payment and channel solutions for Bank of America Merrill Lynch, has high hopes for the deep analytical technology. She told the NACHA payments crowd that the type of data centralization and pattern analysis skill Watson showed in his quiz-show appearance could be useful in accumulating-in real time-the frequent updates from the Office of Foreign Assets Control, and in reducing the number of "false hits" that bog down human analysts pouring over daily exceptions to sanctions rules from the Treasury Department.
Watson's know-how also could be beneficial for making risk-profile assessments in commercial banking, where day-to-day fluctuations in cash flow for corporate customers can leave huge blind spots in exposure for banks.
