Technical objects, like algorithms, exhibit causal capacities both in terms of their internal makeup and the position they occupy in relation to other objects and processes within a system. At the same time, systems encompassing technical objects interact with other systems themselves, producing a multi-scale structural composition. In the framework of fair artificial intelligence, typical causal inference interventions focus on the internal workings of technical objects (fairness constraints), and often forsake structural properties of the system. However, these interventions are often not sufficient to capture forms of discrimination and harm at a systemic level. To complement this approach we introduce the notion of locality and define structural interventions. We compare the effect of structural interventions on a system compared to local, structure-preserving interventions on technical objects. We focus on comparing interventions on generating mechanisms (representing social dynamics giving rise to discrimination) with constraining algorithms to satisfy some measure of fairness. This framework allows us to identify bias outside the algorithmic stage and propose joint interventions on social dynamics and algorithm design. We show how, for a model of financial lending, structural interventions can drive the system towards equality even when algorithmic interventions are unable to do so. This suggests that the responsibility of decision makers extends beyond ensuring that local fairness metrics are satisfied to an ecosystem that fosters equity for all.