I.B.M. Puts Watson to Work in Business

Since its triumph over human “Jeopardy!” champions two years ago, I.B.M.'s Watson has explored a number of frontiers for its artificial-intelligence technology, even inventing new recipes in the kitchen.

To date, medicine has been Watson's most focused foray. Its machine intelligence has read through vast troves of medical texts, research papers and patient records to suggest diagnosis and treatment options at institutions including the Cleveland Clinic, the Memorial Sloan-Kettering Cancer Center and the M.D. Anderson Cancer Center, and for the insurer WellPoint. Yet even its work in health care has had the feel of applied science projects - impressive, but seemingly custom jobs rather than the beginning of a big commercial business.

I.B.M. hopes to take a big step toward mass-market commercialization of Watson on Tuesday, announcing that the technology will be applied to customer service, broadly defined. Its new offering, called IBM Watson Engagement Advisor, is being pitched as a smart assistant whose services can apply to almost any industry, but especially those that receive many customer service calls, like retailing, banking, insurance and telecommunications.

“Watson has a new job - assisting brands,” said Manoj Saxena, the I.B.M. executive in charge of the Watson business unit.

In health care, Mr. Saxena noted, Watson has done “high-value” work for a handful of institutions. The customer-service business, by contrast, is a “high-volume play for us.”

This week's announcement is the next move in an ambitious, longer-term commercialization agenda for Watson. I.B.M. plans to offer the Watson software as the equivalent of an operating system for artificial intelligence applications - much as Microsoft's Windows was the operating system that dominated the personal computer era.

Software developers and start-up companies, according to the I.B.M. game plan, will build software applications that run on the Watson artificial-intelligence engine - and with them, new businesses. In a speech last week at a meeting of the National Venture Capital Association in San Francisco, Virginia M. Rometty, I.B.M.'s chief executive, said, “We'll launch an ecosystem where Watson is a service and you build applications around it.”

The Watson technology can either assist human customer-service representatives or take over some of their tasks. An “Ask Watson” feature allows consumers to ask Watson questions directly from a smartphone or notebook computer and proceed, the company says, in a conversational style, either with typed text or spoken words. (Voice recognition technology would be supplied by third-party specialists, like Nuance, as is the case for Apple's Siri question-answering service.)

The two years of development since the “Jeopardy!” contest, Mr. Saxena said, have made Watson far faster and more compact, whether the cloud service resides on servers in an I.B.M. data center or on a corporate customer's servers.

The original Watson was a question-and-answer machine - with each interaction a separate question and answer. The current technology, according to Mr. Saxena, can engage a consumer in a dialogue. It can listen to and respond to, say, four or five follow-up questions and remember the previous questions a person had.

“That's the difference,” he said, “between a question-and-answer system and a conversational system.”

Those who have been shown demonstrations of Watson's new capabilities sound impressed. Joyce Phillips, chief executive of ANZ Bank's wealth management and private banking group, plans to use Watson to assist financial advisers in coming up with investment, insurance and pension offerings that are better tailored to the needs of individual clients.

ANZ, a major banking group in Australia and New Zealand, is still testing the Watson technology. But Ms. Phillips said: “We're pretty confident that this technology will help transform our business. It's a tool that will make our advisers smarter.”

At the Nielsen Company, the media measurement and market research firm, Randall Beard, global head of advertiser solutions, wants to use Watson to instantly sift through and make sense of all the data advertisers receive about their Web, social media, print and television marketing campaigns.

“You want to optimize campaigns in real time across every marketing channel, to see what people are watching and what are they buying to accurately measure ad effectiveness,” Mr. Beard said. “We think there's a real opportunity here.”

But the Watson opportunity has not yet been tested. Companies so far have signed agreements to develop projects to try out the I.B.M. technology. Out in the wild, when thousands of customers are asking questions all at once, will the promise of Watson's technology prove itself? Will its conversations with people seem smart, as if it is listening and learning?

That would appear to be a tall order. Still, Watson's introduction to the wider world will be far more controlled than Siri's introduction in October 2011. Amid high expectations, fueled by Apple's polished marketing campaign, Siri was instantly subjected to all kinds of messy human questions, on every imaginable subject. Inevitably, Siri often stumbled at first, becoming the subject of ridicule and jokes on late-night television.

But the bar will most likely be lower for Watson in business than it was for Siri in the consumer market. If the basis of comparison for Watson is the current state of voice-automated systems in call centers, the newcomer, analysts say, could well come out a winner.

“People are free-form, and they ask questions in different ways,” said Mary Wardley, an analyst at IDC. “If Watson can learn and offer more personalized answers, it could be a very useful tool in solving customer service problems.”