Bee foraging is complex and observed patterns often deviate from predictions. Much of this deviation can be explained through differences in the characteristics of the flower patch, but not all. We suggest that the position of the flower patch within the landscape can have strong effects on bee patch use, independent of the characteristics of the flower patch, through changes in access to nesting resources and the microclimate. Further, the effects of different habitat configurations are likely to vary among bee groups, depending on flight distances and resource requirements. We surveyed local habitat and plant community characteristics around 100 observation plots in native grassland within temperate savannah and recorded bee patch use by direct observation. Using structural equation modeling, we quantified the direct and indirect effect of plant community and habitat configuration on patch use for all bees, only bumblebees, and only non-bumblebees. Our models explained nearly 50% of the variation in bee patch use, with equal importance assigned to local habitat configuration and the plant community. Flower abundance was the best indicator of patch use, followed by distance to forest edge, site productivity and proportion forest cover. Contrary to previous findings, flower diversity had weak, yet negative effects on patch use after accounting for covariance with other factors. Although all bees most frequently visited plots with high flower densities, bumblebees preferentially used edge habitats, while non-bumblebees were mostly found in productive grassland areas with access to forest habitat. Our results show that small-scale habitat configuration can have direct effects on bee patch use as strong as or stronger than those of the flower patch itself. Further, the exact habitat configuration context that drives bee patch choices is likely to vary among groups. This suggests that habitat configuration should be considered in bee conservation efforts.