Despite progress in the development and evaluation of evidence-based primary mental health interventions in low-income and middle-income countries, implementation and scale-up efforts have had mixed results. Considerable gaps remain in the effective translation of research knowledge into routine health system practices, largely due to real-world contextual constraints on implementation and scale-up efforts. The Southern African Research Consortium for Mental Health Integration (S-MhINT) programme used implementation research to strengthen the implementation of an evidence-based integrated collaborative depression care model for primary healthcare (PHC) services in South Africa. To facilitate the scale-up of this model from a testing site to the whole province of KwaZulu-Natal, a capacity building programme was embedded within the Alliance for Health Policy and Systems Research (AHPSR) learning health systems (LHS) approach. The paper discusses efforts to scale up and embed case finding and referral elements of the S-MhINT package within routine PHC. Data from semistructured interviews, a focus group discussion, proceedings from participatory workshops and outputs from the application of continuous quality improvement (CQI) cycles were thematically analysed using the AHPSR LHS framework. Learning particularly occurred through information sharing at routine participatory workshops, which also offered mutual deliberation following periods of applying CQI tools to emergent problems. Individual-level, single-loop learning seemed to be particularly observable elements of the AHPSR LHS framework. Ultimately, our experience suggests that successful scale-up requires strong and sustained relationships between researchers, policy-makers and implementers, investments into learning platforms and organisational participation across all levels to ensure ownership and acceptance of learning processes.