Crowdsourcing is a popular topic within KM circles. Where knowledge is dispersed, then accessing the knowledge of the crowd may be the best way to gain insight on a particular topic.
Where the crowd as a whole (or at least the majority of the crowd) has experience of a topic, then it can be useful to ask the entire crowd, and take an average answer; for example, the often quoted example of using the crowd to guess the weight of a pig, or the number of marbles in a jar, when the average of the guesses from the crowd is more accurate than the guesses of individual experts. We see this in the "Who want to be a millionaire game, when the audience (crowd) is very helpful on the easy questions.
Where the majority of the crowd has little or no experience of a topic, then they become less reliable en masse. As the Millionare strategy guide tells us - "Once you get through to the later stages the audience becomes unreliable, as on the difficult questions there may still be some audience members who know the right answer, but their contribution will be blotted out by everybody else just having a guess".
In situations like this, you canvas the crowd to find the person with experience, and then you ask them individually for the answer. This is how communities of practice forums work - you ask the crowd a question, and only those with experience and knowledge give you an answer. You find the experts in the crowd, and use their knowledge.
The key to crowdsourcing, as in Millionaire, is knowing when to ask the audience en masse, or when to look in the crowd for a friend to phone