IBM Revamps Computer Network Unit Around Watson AI Platform

Written by hcjc
July 12, 2017

International Business Machines Corp. is revamping its Global Technology Services division, which helps customers run their computer networks, to rely more heavily on artificial intelligence.

(Bloomberg) — International Business Machines Corp. is revamping its Global Technology Services division, which helps customers run their computer networks, to rely more heavily on artificial intelligence.

The new AI-capability will help IBM's customers minimize disruptions such as server outages or switching malfunctions by predicting problems before they occur and automatically taking corrective action, such as brokering additional cloud capacity or rerouting network traffic around bottlenecks, Bart van den Daele, general manager of IBM Global Technology Services in Europe, said in an interview.

The product offering, powered by IBM's Watson cognitive computing platform, will enable the company to maintain its market share in IT network infrastructure management, Van Daele said. New York-based IBM has been struggling to pivot from reliance on older products like computers and operating system software and into higher-growth areas like AI and the cloud. It's facing stiff competition from rivals including Microsoft Corp., Cisco Systems Inc. and Alphabet Inc.'s Google, who have also begun emphasizing artificial intelligence and automation in products that help customers manage their servers and networks.

Recently Microsoft announced a major reorganization of its global salesforce, in part to focus on selling AI-enabled products and services. Since becoming Chief Executive Officer in 2012, Ginni Rometty has increasingly sought to integrate Watson — a suite of different artificial intelligence capabilities linked together — into all of IBM's product offerings. She has done so in part because sales of its traditional products, which included hardware such as mainframes and servers that were physically located in customer's offices, as well as the service contracts to maintain these systems, has been steadily declining. In the first quarter of this year, IBM notched a 20th consecutive quarter of revenue declines.

The move announced Wednesday follows a series of job cuts at IBM's Global Technology Services division last year that were billed as a strategy shift toward more cloud computing and AI operations. But Van den Daele said the announcement about his division's new emphasis on AI-based software "is in no way linked to our work rebalancing. We are focused on driving the next level of innovation in infrastructure. We are not doing it to create labor efficiencies."

For example, instead of employing an analyst to simply stare at a dashboard, trying to interpret network performance data, that person could now be automatically alerted to a slowdown and presented with three options to choose from to get the network running smoothly again, Van den Daele said. The system also has the ability to understand IT helpdesk queries using natural language, meaning that it could take over much of the low-level work that a company's IT support staff have to handle on a daily basis. IBM trained its Watson-based IT infrastructure platform by feeding it data from more than 10 million past incidents, Van den Daele said. The system is now handling more than 800,000 incidents a month.

The new system, which has been piloted with customers including Danish banking group Danske Bank A/S, has already made improvements in network performance, he said. "We saw a significant reduction of server incidents," Jay Steen Olsen, Danske Bank's chief technology officer, said in a statement. He said the IBM platform would "help us act before an incident occurs and move us closer to an integrated, automated and always-on environment.

"Van den Daele said one large food services distributor, which he said he could not name due to customer confidentiality, had been able to reduce critical issues across a network of 4,000 servers by 89 percent and it had reduced the time needed to resolve the remaining issues to just 28 minutes from an average resolution time of 19 hours before the AI-based system was used.The AI-based analytics run on IBM's cloud computing platform, but the underlying data is retained by the customer on whatever network architecture they currently use, van den Daele said.