The explicit inclusion of mental health within the Sustainable Development Goals is a welcome development, borne out of powerful advocacy using public health, economic and human rights arguments. As funding comes on line for scale-up of evidence-based mental health care by task-sharing with primary care, it is time to take stock about care for people affected by severe mental illness (SMI). The existing evidence base for task shared care for SMI provides an imperative to get started, but is skewed towards relatively more affluent and urban populations in middle-income countries where specialist mental health professionals provide most of the care. Randomised, controlled trials and rigorous implementation research on task shared service models are underway which will go some way to improving understanding of the quality, safety, effectiveness and acceptability of more widely generalisable care for people with SMI. A sub-group of people with SMI have more complex and long-term needs for care, with a high risk of homelessness, imprisonment and human rights violations as family and social supports become overwhelmed. Case studies from non-governmental organisations provide examples of holistic approaches to rehabilitation, recovery and empowerment of people with SMI, but rigorous comparative studies are needed to identify the most efficient, effective and scalable approaches to care. Health system constraints are emerging as the over-riding barriers to successful task-sharing, highlighting a need to develop and evaluate chronic care models for people with SMI that succeed in reducing premature mortality, improving wellbeing and achieving better social outcomes. Addressing these evidence gaps is essential if task-sharing mental health care is going to deliver on its promise of promoting recovery for the full range of people affected by SMI.