With the growing availability of advanced demand forecasting models, an important operations question is how to utilise this information to make better resource allocation decisions, especially when facing demand surges. In this work, we design surge routing policies that explicitly incorporate future arrival rates for multi-class multi-pool parallel-server systems with partial flexibility. Our proposed policy is easy to interpret and achieves near-optimal performance. The analysis quantifies the benefit of having access to future arrival rate information. Our development is based on optimal control theory. Time permitting, I will also talk about how to apply what we learn from optimal control to develop numerical algorithms for skill-based routing problems.