PurposeOver the past century, office designers have adapted to changing work conditions across various industries by creating workspaces that cater to the needs of modern workers. As a result, a range of office layouts have been developed, from traditional private offices to open-plan setups. The choice of layout can significantly impact communication, social interaction and overall performance. This study aims to explore the social and mental well-being of employees working in open-plan offices in the public sector in South Africa.Design/methodology/approachThe research was conducted in the Johannesburg Metropolitan Municipality, South Africa, focusing on five governmental departments. Using purposive sampling, employees working in open-plan offices were identified. A structured online interview form with open-ended questions was distributed to all selected employees between May and September 2021, resulting in 54 responses. Thematic content analysis using ATLAS.ti was used to analyze the data.FindingsThe findings suggest that while open-plan offices can enhance teamwork, communication, accessibility and inclusivity, their design can negatively affect employees’ perceptions of social and mental health in the workplace. Specifically, open-plan designs are perceived to compromise privacy, shorten attention spans, lower productivity and increase stress levels. Interestingly, absenteeism was not reported by employees in this study, which contradicts previous research and requires further investigation.Practical implicationsThese findings can be valuable for human resource managers, facilities managers and office space designers.Originality/valueThis research contributes to the ongoing conversation surrounding the impact of open-plan offices on the mental and social well-being of employees in the public sector within the South African context. The contribution of this paper thus lies in the South-African context and the public sector angle plus a more in-depth insight on well-being through an extensive set of 54 interviews.