The past two decades have seen a growing call for researchers, policy-makers and health care providers to collaborate in efforts to bridge the gaps between research, policy and practice. However, there has been a little attention focused on documenting the challenges of dealing with decision-makers in the course of implementing a research project. This paper highlights a collaborative research project aiming to implement the accountability for reasonableness (AFR) approach to priority setting in accordance with the Response to Accountable Priority Setting for Trust in Health Systems (REACT) project in Tanzania. Specifically, the paper examines the challenges of dealing with decision-makers during the project-implementation process and shows how the researchers dealt with the decision-makers to facilitate the implementation of the REACT project. Key informant interviews were conducted with the Council Health Management Team (CHMT), local government officials and other stakeholders, using a semi-structured interview guide. Minutes of the Action Research Team and CHMT were analysed. Additionally, project-implementation reports were analysed and group priority-setting processes in the district were observed. The findings show that the characteristics of the REACT research project, the novelty of some aspects of the AFR approach, such as publicity and appeals, the Action Research methodology used to implement the project and the traditional cultural contexts within which the project was implemented, created challenges for both researchers and decision-makers, which consequently slowed down the implementation of the REACT project. While collaboration between researchers and decision-makers is important in bridging gaps between research and practice, it is imperative to understand the challenges of dealing with decision-makers in the course of implementing a collaborative research project. Such analyses are crucial in designing proper strategies for improved communication and for the utilisation of research projects over time.