Understanding whether pollinator behaviors and species richness drive crop yields is a key area of investigation in pollination ecology. Using sunflower as a study species we describe variation in mean floral visitation times among bee taxa and test how interactions between bee richness and the proportion of bumblebees in localized communities impact seed yield. Seven bee genera commonly visited sunflower including Agapostemon, Bombus, Halictus, Lasioglossum, Megachile, Melissodes, and Svastra. Mean visitation times to sunflower varied across genera and Bombus and Halictus spp. spent the most time foraging on inflorescences, but the number of visits by Bombus spp. was the only parameter associated with increased yields. Experimental pollination deficit reduced seed development and yields, and these effects were stronger in stands of wild-type sunflower in the field compared to a confection variety grown in the greenhouse. Relationships between bee richness and pollination services differed for potted and wild sunflower: when bees had short-term access to potted sunflower, bee richness and relative Bombus abundances were not associated with pollination quotients. When bees had long-term access to wild sunflower, relative Bombus abundances predicted pollination services but were modified by site-level bee richness: as richness increased, the effects of Bombus abundance decreased. Our studies demonstrate that bee species richness is not always a clear predictor of pollination services; instead, our results underscore the importance of specific taxa when species richness is low (here, bumblebees), and show that the effects of bee functional groups important for pollination may be modified by changes in site-level species richness.