The linear, goal-oriented approach to projects that is so popular in management literature is only appropriate if you are dealing with well-defined problems. For projects that address poorly defined problems, however, the principles of classic project management don't work; project managers attempt in vain to maintain a linear approach, even if targets, people affected and framework conditions cannot be determined precisely. We propose a fundamentally different approach based on current organizational theory: to start out with experiments, without predetermined conclusions. Projects are not evaluated by comparing the current status to the target, but rather by assessing whether stagnation has been overcome, conflicts put aside, and shared understanding about new opportunities has been created. Project groups and steering committees are not set up at all. Power "games" are harnessed and put to use, rather than prohibited.