Ever since I first worked in the KM field, I experienced that a lot of pay offs of KM initiatives we saw were actually never originally planned for.
A small technology or tool suddenly became a hit because people understood a different use and implications for it. Quite a few of true KM successes (like the Wikipedia) have been successes despite running against conventional wisdom. That is because KM relies not on technology but on people (who are irrational, impulsive and notoriously difficult to predict) to be successful. Which is why they rarely run how CEOs plan for in the conventional sense.
This view got another validation with this post of Dave Pollard who talks about George Siemens' new upcoming book Knowing Knowledge. As Dave quotes George saying:
Change is a consensual process: If changes (e.g. the use of a new process or website) are mandated by management but don't 'make sense' to those who actually have to effect the change, those people will find ways to not change, and will find workarounds that are effective in spite of the nonsensical mandates of management. Only when there is a consensus that change is valuable will it "take root". The four change enablers in the graphic above actually operate almost like a pendulum: The demand for change (usually from customers, sometimes from management, sometimes from front-line workers' learning and adaptation) precipitates 'affordances' (possibilities, ideas, alternatives and potentials) which, in turn, if they can achieve consensual traction, precipitate structural, systems, and infrastructure change within the organization, which, in turn, finally produce new methods and processes -- different ways of doing things in the organization.