Research question The study investigates the micro-processes of concept generation at ideation stages, on the case of innovation-oriented discussions within online communities of employees. Literature review and research positioning Previous research has already described a variety of approaches and techniques to support collaborative generation of a large number of ideas (Garfield, 2001; Knoll and Horton, 2011; Harvey, 2014). This literature states the importance and interdependence of a cognitive process within individual group members and a social process as group members interact (Knoll and Horton, 2011). Existing innovation studies show that development of collaborative Web platforms facilitates social interaction, enabling different departments to gather knowledge and engage in idea generation (Michaelides et al. 2013). Xiaomi et al. (2013) investigate the role of knowledge management to support Collaborative Innovation Community Capacity Building (CICCB) in order to enhance the effectiveness of innovation within the online community of employees. However, there has been little research that analyzes the micro-process of collaborative concept generation. Taura and Nagai (2012) highlight the concept generation process to be a synthetic one, which is investigated through analogy, blending and integration by thematic relation as a research method (Taura and Nagai, 2012). We propose to empirically investigate this issue in the context of innovation-oriented online discussions. Furthermore, we aim to analyze which sequences of cognitive entities (i.e. groups of words corresponding to elementary ideas) explain the evolution of interactions between the participants and how initial ideas are, successfully or not, developed into concepts. Methodology and empirical material This paper focuses on the nature and dynamics of online discussions at the ideation stage in a virtual idea generation environment. In order to empirically study the micro-processes of concept generation, an action research method (Coughlan & Coghlan, 2002) was employed in this research: through collaboration with a consulting firm, the authors conducted and facilitated four online debates within an international group. The innovation-oriented online discussions were launched in the context of an online community of employees to allow its members to go beyond daily discussions and further explore one trend or opportunity for the group. Our role was to mobilize and stimulate collaborative innovation community capacity (Xiaomi et al, 2013) and creativity of community members during a whole week, in order to support the community ability to generate innovative concepts and ideas for the given topic. To obtain more detailed view of the observed discussions, every interaction and the content of every message posted in the debate were categorized and deeply analyzed through combination of Pena-Schaff and Nicholl’s analytical framework (Pena-Shaff & Nicholls, 2004) and Taura and Nagai design theory framework (Taura and Nagai, 2012) Findings and managerial implications We present three sets of results: firstly, we describe the general dynamics of an online innovation-oriented conversation through its main characteristics (the number of participants, the nature of interactions, the response-time between messages and topics discussed); secondly, we describe the micro-processes of concept generation by deeply analyzing the content of the subsets of messages that contributed to generation. We further analyze the innovative potential of generated concepts for the international group; finally, we propose a framework of participatory roles by correlating the nature of interventions to specific cognitive entities. While limited, this study aims to help researchers and practitioners to better understand the process of concept generation in an online environment and thus, provide the archetypes defined by the patterns of analyzed interactions.