Psychosocial interventions are emerging as potentially important therapies for primary care, partly to fill a therapy ‘vacuum’ and partly because the evidence base for their effectiveness is growing. They can be labour-intensive and their effectiveness depends on the skills of those working with people with dementia. This creates an immediate problem, since the workforce necessary to extend psychosocial interventions beyond innovative schemes does not exist. The existing workforce in health and social care in the UK is already too small to implement all of the changes required by the National Service Framework for Older People and National Service Framework for Mental Health. This has clear implications for the labour-intensive work of dementia care. If the job categories cannot expand as fast as is needed, the tasks of dementia care will have to be redistributed, suggesting that skills will have to be shared and transferred between different disciplines. The question for service commissioners and providers is: how can smarter working be achieved? This article attempts to answer this question with a qualitative study in general practice settings and with specialist informants. A triangulation approach to data collection was used, involving nominal groups, individual interviews and participant observation. We identified five skills that appear key in primary care: pattern recognition; deductive synthesis to reduce uncertainty; dialogue and disclosure; disability perspectives; and case management with shared care. The paucity of understanding of psychosocial interventions across disciplines who offer dementia care in the community is, we suggest, a major problem for those attempting to deploy such interventions in primary care settings. The pervasive tendency to frame the tasks of dementia care in terms of a medical management model brings responses that can undermine the view that people with dementia may in fact have a tractable disability. We use our findings to suggest solutions to this problem.