A redesign process at the farm level may be required for agricultural production systems to evolve in a manner that reduces their environmental and health impacts. This process leads to imagining configurations described as “radical” because they reach beyond the limits posed by the substitution of synthetic inputs by natural ones. An assessment of the possible effects of these configurations on farm functioning and performance is required to inform stakeholders about the advantages of testing and implementing them. This study describes an approach for designing and assessing such configurations that involves researchers, technicians and farmers. Some of these stakeholders can play the role of designers, who lead the redesign process, and/or experts, who provide references and knowledge throughout the exercise. The approach is based on six principles (evaluation, plausibility, precision, flexibility, diversity, iteration) and includes eight steps. Based on a diagnosis of the production context (step 1), some ideas of radical production system are imagined (step 2), which define the kind of experts to be involved (step 3). A farm, virtual or real, then is selected and characterized as a case study (step 4), and the specific objectives driving the farm's redesign process are described (step 5). Scenarios are then designed and characterized (step 6), quantitatively assessed using a simulation tool dedicated to the kind of production system studied (step 7), and compared in order to feed debates between designers and experts on the merits and limits of the various options designed (step 8). Steps 6 through 8 may be repeated as new ideas emerge. This methodology is illustrated with the case of a farm specialized in apple production on which a sheep unit is introduced to reduce pesticide use by ensuring grass management and reducing pest pressure. Two scenarios are designed according to the kind of sheep management. CoHort software was used to assess the two scenarios in