There are a number of factors that help determine how eager people are to share knowledge. Communities that are brought together by technology aren't successful if they don't have a clear, specific, easily measurable business purpose. We also found that there's a really strong "what's in it for me" factor involved. The knowledge-sharing communities that create value for the company also have to create value for the individuals.

One last point: You need strong, engaged leaders to make this work. You need people who will rattle the cages, who will say, "Here's an interesting problem. Does anybody have any insight?"

It seems as if many knowledge-sharing programs have come up short. What parts of this field do companies tend to overestimate the most?

Many companies create elaborate databases of best practices, which I find tend not to work very well -- at least not as a pure database. For example, somebody discovers a nifty way of saving a million bucks and writes up a story for the "Best Practices" feature in our in-house magazine, Agenda. Maybe the idea will get used in exactly the same way in another country -- but maybe not.

Again, it's really not about the technology per se. What generally happens is that people connect at a meeting or at a conference and fill each other in on whatever important issue is on their minds at the time.

Knowledge transfer occurs because the right people happen to meet. We view our databases more as indirect pointers to people than as actual ways of sharing specific ideas.

Can you share some of the obstacles to sharing ideas?

There are plenty of obstacles, most of them having to do with people and organizations. For example, it's hard to share knowledge if you don't have enough time to reflect on what you know or what you need to learn.

Most companies have squeezed almost all of the reflection time out of their business processes. People don't have the time to think about what they're going to do next, let alone who they should talk to about it. People get extremely task focused: "What's the least amount of work I can do to get this done?"

So we have built a specific process into the project-management techniques that we use at the company. It calls for bringing in others to consult with you, and to review what you've done, to see if they have any suggestions or advice.

What can leaders do wrong in this process?

It's easy for leaders to screw up a community of people working together. Communities tend to disappear when you try to control them, or even when you try to "help" them. Organizations are living organisms, not engineering artifacts. If you take an organization apart, spiff it up a little, and then put it back together, it usually doesn't work the same way.

Fara Warner is a Fast Company senior writer based in Silicon Valley. Contact John Old by email (oldjm@texaco.com).

Sidebar: Smart Mail

It's hard enough to think great thoughts -- let alone to capture them for a knowledge database. That's why David Gilmour, the founder, president, and CEO of Tacit Knowledge Systems Inc., in Palo Alto, set out to create applications that capture expertise as it gets created. Tacit's software and server products scour email, documents -- anything digital -- to build an ever-changing knowledge bank.

Gilmour calls it "expertise automation." KnowledgeMail keeps track of when people stop writing about a topic, as well as how often certain words or phrases are used and how important those words are within the document. To illustrate the way KnowledgeMail works, Gilmour uses a law firm as an example.

"Let's say that your case has been reassigned to a judge whom you know nothing about," he explains. "In the past, you'd wonder, 'Who can I call to find out what the judge is like?' But that takes time -- as well as knowledge of all the lawyers in your firm and which judges they interact with." With Tacit, you type in the judge's name, and all of the references tied to that name pop up. Not only does the software give you people who have tried cases before that judge -- but it also gives you details such as how often, over what span of time, and what type of case was tried.

Contact David Gilmour by email (davidg@tacit.com).