In order to obtain precise theoretical predictions for collider experiments, it is often necessary to evaluate higher-order corrections in QCD. The calculation of these effects is complicated by infrared singularities which need to be removed before the problem can be treated using numerical methods. In this talk, I will report on the development of a framework to automate the local subtraction of infrared singularities beyond next-to-leading order (NLO). After introducing overlapping singularities, I will discuss common ideas and tools of subtraction schemes and outline a new possible strategy to remove implicit divergences in the real-radiation phase space.