Increasing the levels of ecosystem services that contribute to agricultural production (ES) is a challenge for the sustainability of agricultural systems. Agricultural advisors lack low-data operational approaches for assessing ES and knowledge to support development of ES-based systems. To fill this gap, we developed an approach that assesses relations between characteristics of agroecosystems and the ES they offer: pollination, pest, weed and disease control, soil structuration, nitrogen and phosphorus supply to crops, water retention and control of erosion. We distinguished four dimensions of ES: potential capacities, real capacities, levels actually used by farmers and dynamics of the natural capital that supports ES provision. We assessed them with a low-data in-dicator-based method at the agroecosystem level. It provided a score for (i) the quality of the agroecosystem's spatiotemporal configuration, (ii) positive or negative modulations in ES expression caused by agricultural practices, (iii) the farmer's strategy for using ES and (iv) four components of natural capital dynamics (soil quantity, soil organic matter, phosphorus cycling and the biodiversity that supports ES). We demonstrate the interest of the approach by applying it to 34 contrasting agroecosystems in France and subsequently identifying five agriculture models. Analysis of this case study identified several ways to attain high-yield agroecosystems based on anthropogenic inputs, ES or both. We discuss strengths and possible improvements of our approach and highlight key knowledge gaps to examine in future studies.