Ecosystem services are defined as benefits produced by ecological communities, supporting human welfare. Because sustainable agriculture relies on such ecosystem services, finding the optimal management – which optimizes both the surface dedicated to human activities and the delivery of ecosystem services – is particularly critical. Ecosystem services heavily depend on the presence and activity of organisms, especially ecosystem engineers. In order to find the proportion and the spatial aggregation of exploited areas that optimize an ecosystem service, we developed three complementary metapopulation models of a keystone species in an exploited landscape. We considered both anthropic and ecological constraints, by modelling the simultaneous management of two variables: the yield of human activities and the ecosystem service provided by the metapopulation. We also investigate how this optimal management can drive the metapopulation close to extinction, and how two key ecological traits of species – population growth and dispersal rates – can mitigate such extinction risks. The two spatially implicit metapopulation models show that the optimal management is a trade-off, benefits often being optimized for intermediate surfaces of exploitation. This optimal surface depends on the ecological traits and on the degree of disturbance incurred by human activities. Spatially explicit simulations suggest that optimal management is further improved when the spatial distribution of human activities is fragmented.