The debate on future global food security is centered on increasing yields. This focus on availability of food is overshadowing access and utilization of food, and the stability of these over time. In addition, pleas for increasing yields across the board overlook the diversity of current positions and contexts in which local agriculture functions. And finally, the actual model of production is based on mainstream agricultural models in industrialized societies, in which ecological diversity and benefits from nature have been ignored or replaced by external inputs. The dependence upon external inputs should exacerbate the negative impacts on the environment and on social equity. Strategies to address future global food security thus require local innovation to increase agricultural production in a sustainable, affordable way in the poorest regions of the world, and to reduce the environmental impact of agriculture and its dependence on non-renewable resources. Ecological intensifi cation, the smart use of biodiversity-mediated ecosystem functions to support agricultural production, is portrayed as the most promising avenue to achieve these goals. Here we fi rst review examples of ecological intensifi cation from around the world. Functional diversity at plant, fi eld and regional scales is shown to hold promise for reducing pesticide need in potato production in the Netherlands, increasing beef production on the pampas and campos in south-east South-America without additional inputs, and staple crop production in various regions in Africa. Strategies range from drawing on high-tech breeding programs to mobilizing and enriching local knowledge and customs of maintaining perennials in annual production systems. Such strategies have in common that larger spatial scales of management, such as landscapes, provide important entry points in addition to the field level. We then argue that the necessary innovation system to support transitions towards ecological intensifi cation and to anchor positive changes should be built from a hybridization of approaches that favour simultaneously bottom-up processes, e.g. developing niches in which experiments with ecological intensifi cation develop, and top-down processes; changing socio-technical regimes which represent conventional production systems through targeted policies. We show that there are prospects for drawing on local experiences and innovation platforms that foster co-learning and support co-evolution of ecological intensifi cation options in specific contexts, when connected with broader change in the realm of policy systems and value chains. This would require dedicated system innovation programmes that connect local and global levels to sustainably anchor change towards ecological intensification.