In French Atlantic marshland, farming systems are operating in a complex socio-pedo-climatic environment characterized by grazing livestock systems on wet meadows but also cropping systems on drained heavy clay soils. Managing farming systems supposes to manage trade-offs between key ecosystem functions of wetlands : food production, biodiversity conservation, water quality regulation and flood buffering. Therefore, switching from conventional to agroecological cropping systems is a major challenge in these areas. This transition is constrained by the limited numbers of days with adequate soil conditions (constraint exacerbated by the greater climate variability, especially high levels of winter rainfall events). It is also due to weed species adapted to the hydromorphic conditions of these environments. The aim of this study is to test a theoretical framework for the progressive design of farming systems (Meynard et al, 2023), applied to an entire organic farming system seeking to diversify its food production. Here we present the learning feedback loop during the implementation of this farming system after 5 years of cultivation. We focus on analyzing the pathway to move from the initial situation to the targeted one.