At the present time, the increasing precision of LHC measurements has made Next-to-Next-to-Leading (NNLO) accuracy in theoretical predictions mandatory. Despite the impressive advancements achieved over the last years, a final strategy to tackle NNLO calculations is still missing. In this respect, generalized unitarity techniques, which were at the heart of the NLO-revolution, can offer a promising alternative at multiloop level as well. In this talk, I will review the role of unitarity in the reduction of higher-order scattering amplitudes to master integrals and I will discuss some recent ideas, based on maximal unitarity, for the evaluation of master integrals beyond multiple polylogarithms.