Both of the writers of this contribution are social researchers in the UK civil service, working alongside civil service colleagues developing policy proposals. They are based in the central government department that works with local government, planning, neighbourhoods and housing, as well as encouraging decentralisation, community rights and devolved decision making. As social scientists operating within a politically oriented policy-making organisation, they are well placed to trial new ways to bring multiple forms of expertise into policy. These contributors present an overview of the Neighbourhood Community Budgets pilot programme that aimed to devolve decision making about local public spending down to neighbourhood level. The programme can be seen as striving towards integrating some elements of co-productive policy making, such as the emphasis on community intelligence and co-design of policy, into institutions. Its programme was also shaped by what happened in local areas in a series of iterative processes from which central government learned as the programme developed. Ambitious in scope, the programme showed how there was enthusiasm for a change at a local level, which was tempered but not diminished by all too familiar barriers. It also illuminates how policy researchers and evaluators are attempting to live out some key principles of more iterative and incomplete policy design.