There are some problems with this. First, it assumes you know up front the best way to solve the problem. Second, it assumes that the problem doesn't ever change. Third, it completely ignores a suprisingly common issue: the tradeoff. By this I mean: sometimes modularity/flexibility in one area requires being unflexible in another. Guessing wrong requires a big refactor. There are good arguments for both -- sometimes you just have to do one and find out if you were right.
The problem comes when management says "how come you didn't do it right the first time?" and refuses to accept that thier three sentence change of spec requires serious redesign.