Purpose: Clinical supervision is a crucial workplace practice for professional learning and development. Research is needed to investigate in detail what happens in supervision to understand how this practice contributes to learning. This paper aims to examine how professionals work with knowledge and navigate epistemic challenges in working with problems of practice. Design/methodology/approach: Three pairs of psychologists audio-recorded five consecutive supervision sessions and were interviewed twice during that time. Analysis considered supervision as a site of emergent learning, focusing on what was discussed and how problems were worked on, whether as epistemic objects (open-ended, aimed at generating new insights) or by using an approach to knowledge objects that focused more directly on what to do next. Findings: One pair consistently adopted an epistemic object approach, while another was consistently more action-oriented, focused on knowledge objects. The third pair used both approaches, sometimes expanding the object with a view to gaining insight and understanding, while at other times focusing on next steps and future action. Originality/value: To the best of the authors' knowledge, this study is the first to study clinical supervision in terms of how knowledge work is done. Foregrounding the epistemic dimensions of supervision, it reveals previously unexplored but consequential differences in how knowledge is worked with and produced as supervisory pairs discuss complex issues of practice.