There is a clear need for collaboration within organisations, and the rollout of collaboration tools will bring many benefits. What is not widely recognised, however, is that the unmanaged spread of collaboration tools can work against knowledge sharing.
The basic flavor of the problem is that too-good collaboration technologies can create silos amongst the new connections that are formed within the tool, limiting the ability to share knowledge across these new networks of people.
More specifically, James talks about the normal behavior of people to share knowledge within their local network, whether that is people in the nearby cubicles or on the org chart or in the collaborative space. People are likely to use "local" terminology and context that keeps people out, if they not part of that local community. This happens even if the technology itself is open to outsiders. If the collaboration tools lock people out or otherwise keep "knowledge" locked into the space, then this could make sense.
The solution to this problem? While the article doesn't dive into the details, I suspect mechanisms to help are going to include creating both technological and human interfaces between communities and from the local communities into the larger network. Technically, there needs to be a way to promote "finished" articles and probably a process around deciding what looks like finished and how to publish them for the organization. This fits under content management. On the human front, a community monitor of some sort needs to help cross boundaries of this sort. Maybe they need to be required to sit in the discussions of some other communities, just to hear what is happening. This fits perfectly for those curious types (like me).