Research and Technology Organizations (RTOs) are agents which conduct ‘extra-university research’ at the boundary between industry and science, acting as an intermediary between the two. Their expected impact is primarily industrial: they are supposed to support and enhance the competitiveness and innovation capabilities of industry. RTOs position of intermediary between science and industry is delicate, since their role is neither to substitute for the industrial partners who conduct their NPD projects and develop innovations, nor to substitute for the conduct of basic research by universities. In particular, one major pitfall for RTO pointed out by the literature is the risk that the outputs of RTOs activities are too research-oriented and fail to meet industrial needs. In this context, the literature tends to recommend that to remain manageable, the intermediation role of RTO between science and industry should be limited to situations of low/moderate uncertainty. The resulting impacts will be modest, but under control for the RTO and highly-valued by industrial firms. This paper focuses on the French IRT (Institute for Technological Research) SystemX that seems to stand out within the portray of RTOs. Created in 2012, SystemX conducts collaborative research projects aimed at accompanying and accelerating the transformation of industrial sectors in the face the digital transition. Internally, SystemX has intuited that it is inventing a new form of science-industry coupling. Thus, this paper involves mobilizes the theoretical frameworks of design theory, with the aim of characterizing SystemX model of action within its industrial and scientific ecosystem, and testing the hypothesis according to which SystemX has developed capabilities to manage situations the level of uncertainty of which goes beyond moderate uncertainty. The study reveals that SystemX manages science-industry couplings which do not only involve moderate uncertainty, but also unknowns, associated with which there is a high potential of ‘double impact’ (simultaneously a scientific and an industrial impact). SystemX is able to articulate the action model associated with these science-industry couplings in the unknown with three other action models (involving lower degrees of unknown). An in-depth case study focusing on one SystemX project shows that the implementation of the four action models and their articulation requires methodical and progressive processes of structuring and creating knowledge: these processes allow the construction (and not the simple identification) of common locks and new disciplines in the unknown. These could be seen as first insights regarding the conditions required for ‘double impact RTOs’.