Microsoft\'s Plan to Sell Answers

Microsoft may be looking at its ventures into machine learning as providing more than just better features for products. They may be a way to transform itself, and many other companies too.

I wrote an article in The New York Times this week that said Microsoft's 100 top managers had been told to think of ways their businesses could incorporate machine learning - which is the engineering of algorithms that analyze textual, behavioral and sensor data, and then try to form patterns and predictions. It is at the heart of the Bing search engine, but has many other potential applications, including better logistics planning, the organization of a calendar or the design of a marketing campaign.

Craig Mundie, Microsoft's chief research and strategy officer, sees machine learning as a potential means to create entirely new businesses within the company.

Microsoft, he said, “historically sold the software itself, and now in the cloud and services business we'r e not selling you the software, we're selling you a service that the software creates.” By adding machine learning, he explained, “what you could sell to people are answers to things.”

Those answers could be uniquely valuable if they drew from Microsoft's own data resources, culled from assets like Bing, the Windows Azure cloud computing platform and Skype, he said.

“We might have unique capabilities in terms of I.Q., algorithmic capability and infrastructure that the average company wouldn't otherwise have,” Mr. Mundie said. “We may be able to produce an answer to a question that people otherwise can't produce. And to the extent you can do that, it could become another product.”

Microsoft is already headed toward the answers business by increasing the capabilities of tools it offers. The next version of its Excel spreadsheet program, the world's most popular data analysis tool, will be able to incorporate large volumes of new data obtained in real time, from both outside sources and the data sold on Microsoft's Azure Marketplace.

“There is not going to be another enterprise application created from this day forward that does not have some amount of intelligence that's been gained by reasoning over large amounts of data,” said Satya Nadella, president of Microsoft's server and tools business.

“Right now, most enterprises will come to us and say, ‘We have a risk management problem,' or, ‘We have an exploration simulation problem,'” Mr. Nadella said. Microsoft could, he said, maintain a set of algorithms in the Azure Marketplace that customers could rent as needed to solve these problems.

The next step, offering people insight into the type of data they might want, will involve educating them in the tools and skills they should draw from. This may be among the greatest challenges for the business, given the typical mind-set of most businesses.

“There is a class of people called enterprise data warehouse zealots,” said David Campbell, who leads development of Microsoft's SQL Server, which is the basis for many other Microsoft products, like SharePoint. “They have this notion of one version of the truth, that nothing is of consequence unless it's in their enterprise data warehouse.”

In the world of Big Data, however, truth tends to be contingent, based on the needs of the moment and the data at hand. Data on the speed with which an online order is filled means one thing to a marketing person, looking to maximize customer satisfaction, another to a security professional trying to stop fraud, and a third to a network engineer trying to balance resources. “In this new world,” Mr. Campbell said, “there are multiple versions of the truth.”

Different versions will come from “the customer's information, with information that we have,” he said. “We're one of the few entities on the planet that has a fresh snapshot of the Web, and represents all sorts of online and social activity every day. To be able to join these things in interesting ways to answer a bunch of questions is very, very interesting.”

Left unanswered is whether Microsoft's managers, who are largely hard-core engineers schooled in certainty, can adapt themselves to the products that Mr. Mundie and Mr. Campbell see the company selling. Or for that matter, whether they can accept an uncertain world, bounded mostly by contingency.