In recent years, a growing number of workplaces have implemented activities where employees are invited to voice problems and suggest initiatives directly to their managers in a group setting. Such direct group-based voice activities (DGVAs) are typically inspired by human resource management and production improvement techniques, and they are claimed to have a number of positive effects for both the organizations which host these activities and for their employees. However, others have questioned whether they provide employees with a reasonable opportunity to influence their working conditions, or if they instead mostly assign new responsibilities to the employees and promote overcommitted employee identities. This ambivalence regarding the activities is reflected in how the circumstances regarding voice in the workplace are sometimes described as messy and paradoxical. The aim of this dissertation is to understand an important aspect of how employees can influence their workplace through DGVAs, specifically how the participants construct change initiatives which can improve the employees’ working conditions. To this end, the dissertation presents an interaction-focused perspective on voice based on ethnomethodological conversation analysis, a perspective which addresses various shortcomings of the dominant research perspectives on voice. For example, substantial attention has been paid in the voice literature to how individual employees make choices about what messages to convey through voice and whom to address, especially in studies which have applied a psychological lens. However, in DGVAs, voicing a problem or a suggestion to the other participants is only the first step of a longer process towards potential consensus about which initiatives to implement, and the social and interactional mechanisms which underlie this process are not well understood.