The concept of Collective Leadership (CL, broadly speaking leadership within groups) is difficult to define and detect empirically. A promising avenue for detecting CL focuses on discursive approaches based on group interaction and ‘turning points' in the discussion, where participants concur on the need for action. In this methodological paper, we present: a novel NLP-task for the detection of CL and a novel generative AI experimental architecture to support this task. To our knowledge, this research is the first to combine NLP and leadership theories with strong organisational process models. These models are linked using a formal notation applied to board text data, and cemented in a generative AI pipeline with the latest methodologies for in-context learning, reducing the need of costly manual annotation. Forthcoming research will provide an annotated dataset.