The fascinating collective behavior of bacteria, bird flocks, fish schools or insect swarms, is often modeled using interacting, self-propelled stochastic agents. Such individual-based models, offer a natural framework for the mathematical description of such systems. However, due to their complexity, they rely often on large-scale numerical simulation. In recent years, significant progress has been made to derive systematically coarse-grained theories from microscopic dynamics, which describe the system at large scales and are open to analytical investigation. However, the corresponding derivations rely often on some sort of mean-field assumptions, which neglect fluctuations due to finite number of individuals. Here, we will discuss the role and properties of this fluctuations, which become important in small systems or at mesoscopic scales, as well as possible ways to account for them in coarse-grained theories.