Search OnSearch On

Kaiser spent $28,000 on a Google appliance to search an index of 150,000 documents. In the three months since the application went live, the number of searches has risen steadily. The low price meant Hochhalter didn't need budget approval, and there's little drain on IT staff time for maintenance. Hochhalter hopes the IT department will establish Google as a standard. "They're gonna have to pry my cold hands off the keyboard to take this off the site," he says.

Bank One and Kaiser understand something many other companies miss, says Matthew Berk, a senior analyst at Jupiter Research: Employees are a lot like consumers when it comes to search. "The thing businesspeople get wrong is that they treat their employees like a captive audience," Berk says. "They're as fickle as consumers, and they have more hang-ups about search."

But is a search engine alone enough for sophisticated knowledge work? While Google satisfied Bank One and Kaiser, it wouldn't work for KPMG's U.K. division. The business, which provides tax, assurance, and financial-advisory services, wants to create search-tendency profiles to help employees connect with others conducting similar searches, so they can share ideas and results. "We want to move from connecting people to content, to connecting people to people," says Iain Simpson, U.K. information architect. "Had we not been thinking about two years from now, I could easily have been seduced by Google's capabilities."

Last year, the division deployed Verity's K2 software. The knowledge-management and information-infrastructure technology combines sophisticated taxonomy and classification with so-called social networking that connects employees who are searching for similar things and remembers what users have searched for in the past.

In the past 10 months, the U.K. division has rebuilt its back-end information infrastructure to take better advantage of K2's taxonomy and categorization capabilities and redesigned the system's user interface. The next step will be to hook into nearly 50 KPMG intranets around the world. The Verity deployment is for 12,000 users. The rebuilt back end involved completely removing a Verity information server that predated the K2 software, rewriting all the code, and redesigning the server environment. The interface was rewritten to better align with Internet search engines and give users more choices for how they conduct their searches, such as by keyword, date, author, or department.

Other companies have turned to niche vendors that offer specialized capabilities they can't get from Google's simple keyword approach or Verity's taxonomy-intensive orientation. Gateway Inc. wanted to give employees useful search results using general terminol- ogy. It also wanted customers, who aren't always sure what they need to ask, to be able to use the same system. In mid-2002, the computer vendor deployed a natural-language tool from iPhrase Technologies Inc. to 6,500 Gateway employees in 12 call centers and 220 retail locations, with access to 700,000 product-documentation files and issue-resolution descriptions. Since Gateway went live with the system, support technicians have been able to get faster access to more relevant search results, which has improved service, says Mark Notarainni, director of E-support and technology for Gateway's customer-support group.

At the government solutions unit of IT services firm EDS, VP of knowledge services Joe Williamson is dealing with the same problems as many managers: mushrooming content, the need to find expertise quickly, and the desire to get more-relevant results. Added to that is the fact that the 5,700 knowledge workers he serves are scattered around the world and may play different roles on different days.

Williamson wants a search tool that will let employees build profiles and get results based on the jobs they're performing on a given day. A consultant working on a sales opportunity one day and a research-and-development project the next would get different results from the same query depending on what job she's doing that day. Williamson also wants the software to detect relationships among documents in different languages without running them through a translation engine.

While he hasn't found a tool that can do it all, Williamson is using Recommind Inc.'s MindServer, which understands that a single term can have multiple meanings. That capability makes it possible to deliver different sets of role-specific results for identical searches. Williamson says that MindServer is also excellent at categorizing a large volume of data, as it did for the National Library of Medicine's 11 million documents.

Of course, that still leaves EDS and thousands of other companies far short of instant neural downloads. But delivering faster, more-relevant results is a good start. If nothing else, it will help close the employee-competency gap that Ford's Sketch would like to eliminate. And that's the real value of the technology, Sketch says. "We think search is the future of learning."