Building A Better Skunk Works

Starting out in 2000, "we were probably a little late to the party around high- performance computing in life sciences," Kovac recalls, "so we needed some significant wins." IBM wasn't really active in early genomic projects, but it seized the chance to use its database software for newer and even bigger studies of proteins. From there, Kovac helped recruit "marquee engagements" with Johns Hopkins University and the National Institutes of Health. By 2003, the EBO was a $1 billion a year business, with half its revenues coming from small and medium-sized companies that IBM had never talked to before. Kovac was rewarded when IBM folded all its other health-care clients into her division, giving her a $4.8 billion-a-year fiefdom.

Now Kovac is also leading a new EBO focused on a more specific concept that IBM calls "info-based medicine." The idea is to compile and tap huge databases and apply computational power to produce better and more-personalized therapies for patients. One of the pilot users is the famed Mayo Clinic, which has put the medical records of its 4.4 million patients into an IBM database that can be accessed by any of its 2,400 physicians. They use it to run queries about cases similar to ones they're dealing with at present. In theory, that perspective should help doctors tailor treatment to the current patient while minimizing harmful side effects. This year, the business will produce hundreds of millions of dollars in revenue, and before long it will probably break out into the multibillion-dollar range.

Kovac says that the ideal EBOs are businesses where the "tipping points" are two to three years away, so now is the crucial time to prove the concept, forge the key relationships in the marketplace, and position yourself as a leader before the boom.

The EBO program was launched by Gerstner and accelerated by his successor as CEO, Palmisano. The point man all along has been Harreld, who continues to meet monthly with each EBO leader. Often the pow-wows are held in the conference room adjoining Harreld's private office across the hall from Palmisano's. On a side table there are two large glass jars -- one labeled valium (which actually contains pretzels) and the other novocaine (which is full of jelly beans). The jars are a holdover from the tense meetings at IBM in the 1990s, when Gerstner was jettisoning businesses in his successful turnaround effort. The table also displays talking figurines of the comedians Abbott and Costello, which reenact part of the duo's famous "Who's on First?" routine, yet another one of Harreld's tools to goad people to sort through the issues.

Gerstner himself recruited Harreld to IBM in 1995, and in many ways their backgrounds are similar. Both were management consultants who became executives in big food companies -- Gerstner at McKinsey & Co. and later Nabisco, Harreld at the Boston Consulting Group and then Kraft. After a stint as chief information officer there, Harreld taught business courses at Harvard and Northwestern before Gerstner called him and said, "You have to come back to the real world."

Now, looking for emerging business opportunities, Harreld spends a lot of his time talking with Silicon Valley venture capitalists such as John Doerr and Vinod Khosla at Kleiner, Perkins, Caufield & Byers, and he keeps a staff in northern California as a listening post. He says that venture capitalists don't need IBM's money -- they already have billions waiting to invest -- but both parties want to understand each other's interests. "We look at venture capital for 'headlights' about what's going on," he says.

"You want to celebrate failure because you learn something. You need some level of security to say, 'I screwed it up,' and be comfortable that you won't be fired."

Harreld's record with EBOs at IBM is remarkable. His hits include IBM's $2 billion Linux business, which charges for consulting about the free software; a digital media EBO, which helps companies manage video, audio, and images, has grown into a $1.7 billion business in just three years. Some of his most recent efforts are focused around geographic areas -- he has targeted China, India, and Brazil with EBOs. The three markets combined represent an estimated $60 billion opportunity for info-tech providers. He's also looking at Russia and Eastern Europe.

But not every one of Harreld's EBO initiatives has succeeded. A couple of doomed EBOs tried to capitalize on the Internet boom just before it busted in 2001. There was, for example, the "network processor," a chip IBM planned to sell to networking hardware companies like Cisco. Problem was, there were few such companies left (other than Cisco) after the crash. When a pilot doesn't work, Harreld quickly kills the EBO and finds another important position at IBM for the erstwhile leader who took the risk: "You want to celebrate failure because you learn something. It's harder to do that early in your career. You need some level of security to say, 'I screwed it up,' and be comfortable that you're not going to get fired," he says.

Palmisano's vision is for the spirit of EBOs to spread throughout IBM, and that's what motivates the people who've been involved early on in the program. "Doing something like this within a large company such as IBM is not like being an entrepreneur, but it is entrepreneurial compared to the rest of the environment," says Kovac. "I felt if we could do this right, it would change the ability of this company to move from ideas to markets. Our journey isn't over, but it truly is a profound change."

Alan Deutschman is a Fast Company senior writer based in San Francisco.