A crucial aspect of the complexity approach is how interacting elements produce aggregate patterns that those elements in turn react to. This leads to the emergence of aggregate properties and structures that cannot be guessed by looking only at individual behaviour. Explicitly considering how heterogeneous elements dynamically develop their behaviour through interaction is a hard task analytically, the equilibrium analysis of mainstream (neoclassical) economics being a not neutral shortcut. On the other hand, explicitly considering the dynamics of the process started to be a feasible alternative only when computer power became widely accessible. The computational study of heterogeneous interaction agents is called agent-based modelling (ABM). Interestingly, among its first applications a prominent role was given to economic models, although it was quickly found of value in other disciplines too. Goal of this lecture is to motivate the use of the complexity approach and agent-based modelling in economics, by discussing the weaknesses of the traditional paradigm of mainstream economics, and then explain what ABM is and which research and policy questions it can help to analyse.