The remarkable efficiencies of solar energy conversion attained by photosynthetic organisms derive partly from the designs of the light-harvesting apparatuses. The strategy employed by nature is to capture sunlight over a wide spectral and spatial cross section in chromophore arrays, then funnel the energy to a trap (reaction center). Nature's blueprint has inspired the conception of a diversity of artificial light-harvesting antenna systems for applications in solar energy conversion or photonics. Despite numerous, wide-ranging studies, truly quantitative predictions for such multichromophoric assemblies are scarce because Forster theory in its standard form often seems to fail. We report here a new framework within which energy transfer in molecular assemblies can be modeled quantitatively using a generalization of Forster's theory. Our results show that the principles involved in optimization of energy transfer in confined molecular assemblies are not revealed in a simple way by the absorption and emission spectra because such spectra are insensitive to length scales on the order of molecular dimensions.