The paper presents and empirically applies a neo-Schumpeterian model of innovation capable of studying interactions between service providers, patients and policy makers, and how these complex interactions determine the timing, direction, and success of innovations in the public sector. The model is tested using a case study that traces the introduction and development of ambulatory surgery in a Spanish hospital. The multi-agent model applies the ideas of Schumpeter to services, encompassing Schumpeter's five types of innovation, and re-introducing the policy-maker as a key agent in the innovation process. The model has a number of advantages over previous, reduced form models. First, it can analyse the interactions between the economic, social and political spheres that make up the complex selection environment of innovations. Second, it captures the recursive impact of radical innovations on agents’ competences and preferences, and their relative power. This brings politics, power, and rhetorical persuasion to the fore. Third, it provides an improved set of definitions for radical and incremental innovation. These are not only important for understanding the sources and drivers of innovation, but also for the accurate measurement of innovation.