Innovation is central to the strategic orientations of many agronomic research institutes. Little attention has been paid, however, to the links between knowledge production and design processes, defined as processes resulting in the creation of new objects in pursuit of specific goals. Our aim was to analyze the conditions and specificities of the production of scientific knowledge on agro-ecosystems through design processes. Drawing on design theory, we carried out a cross-analysis of nine research projects that included design processes and produced innovative objects (for example blending rules for variety mixtures in low-input crop management routes). These projects were managed by researchers from a range of disciplines (agronomists, geneticists, crop physiologists, and ecologists), and varied in their duration (from three to 15 years) and scale (from plot to landscape). We combined semi-structured interviews with these researchers and the analysis of various documents (scientific papers, PhD theses, technical publications, and research projects or reports). Our findings show that in all case studies, original and general scientific knowledge on agro-ecosystem functioning was produced at various stages throughout the design processes. The originality of this knowledge lies with the new representations that emerged, either of the agro-ecosystem processes at stake, or of farmers' practices. We show that these representations were formed gradually, through successive iterations of both refined formulations of the design target and new knowledge produced, required for its design. Finally, our results highlight the role of confrontation with real-life situations (particularly through agronomic diagnosis or experiments) in the evolution of these representations. Engaging in design processes can thus be seen as a research practice that leads to the production of original knowledge, allowing for a greater diversity of actors' ways of knowing to be taken into account.