Enhancing floral resources is a widely accepted strategy for supporting wild bees and promoting crop pollination. Planning effective enhancements can be informed with pollination service models, but these models should capture the behavioural and spatial dynamics of service-providing organisms. Model predictions, and hence management recommendations, are likely to be sensitive to these dynamics. We used two established models of pollinator foraging to investigate whether habitat enhancement improves crop visitation; whether this effect is influenced by pollinator foraging distance and landscape pattern; and whether behavioural detail improves model predictions. The more detailed central place foraging model better predicted variation in bee visitation observed between habitat types, because it includes optimized trade-offs between patch quality and distance. Both models performed well when predicting visitation rates across broader scales. Using real agricultural landscapes and simulating habitat enhancements, we show that additional floral resources can have diverging effects on predicted crop visitation. When only co-flowering resources were added, optimally foraging bees concentrated in enhancements to the detriment of crop pollination. For both models, adding nesting resources increased crop visitation. Finally, the marginal effect of enhancements was greater in simple landscapes. Synthesis and applications. Model results help to identify the conditions under which habitat enhancements are most likely to increase pollination services in agriculture. Three design principles for pollinator habitat enhancement emerge: (a) enhancing only flowers can diminish services by distracting pollinators away from crops, (b) providing nesting resources is more likely to increase bee populations and crop visitation and (c) the benefit of enhancements will be greatest in landscapes that do not already contain abundant habitat.