This article from Brian Kennemer has been waiting for me to comment on it. KM and PM: The Redheaded Step Children of all Organizations?
KM and PM initiatives fail because the people involved have not yet decided that KM and PM are good for them. Something about the sharing of knowledge and changing the way projects, resources and work are managed makes them uncomfortable and they do they natural thing: they resist. They find things wrong with the system. They don't use it. They poke holes in it and try to kill it. The whole range of defense mechanisms come into play. The most common is to just not use it. It is not an active hatred that kills the new system or process but a slow passive death from starvation.
On top of that, Brian talks about the psychology of KM and PM (and other) "systems" that make their appearance known by the fact that they are software.
People see lots of problems being "solved" with software so they get the idea that software is what actually solved the problem. In some cases like in database systems making it easier to find information this is mostly true but the general misconception stands: Software fixes problems.
Software is only one component of an effort to create change and operate a new way. Unfortunately, many change efforts place so much emphasis on the software that the greater changes in the way one conducts business are lost to the altar of the computer. When we successfully implemented Critical Chain Project Management at Pharmacia, the software was only one component of a larger change. Even better, when people started hearing new kinds of questions from management, they asked us for guidance on how to use the software to answer those questions because their old tools did not provide the right information.