Work-related stress is a global problem causing suffering and economic costs. In Sweden, employees in human service occupations are overrepresented among persons on sick leave due to mental health problems such as stress-related disorders. The psychosocial work environment is one contributing factor for this problem, making it urgent to identify effective methods to decrease stress at the workplace. The aim of the study is to evaluate a participatory intervention to improve the psychosocial work environment and mental health using an embedded mixed methods design. The study is a controlled trial with a parallel process evaluation exploring fidelity and participants’ reactions to the intervention activities, experiences of learning and changes in behaviours and work routines. We collected data through documentation, interviews and three waves of questionnaires. Our results show small changes in behaviours and work routines and no positive effects of the intervention on the psychosocial work environment nor health outcomes. One explanation is end-users’ perceived lack of involvement over the process causing the intervention to be seen as a burden. Another explanation is that the intervention activities were perceived targeting the wrong organisational level. A representative participation over both content and process can be an effective strategy to change psychosocial working conditions and mental health.