The large instantaneous luminosity delivered by LHC to the CMS experiment in Run 2 is an important opportunity for the study of rare B-hadron decays, like $\mathrm{B}^0_{(s)}\to\mu\mu$ and $\mathrm{B}^0\to\mathrm{K}^{*0}\mu\mu$. The main drawback of this luminosity increase is that the trigger selections need to be tighter, in order to fit the rate of recorded events in the limits imposed by the available computer resources. This constraint imposes a challenge in the development of new trigger tools and algorithms, optimized to keep high acceptance on signal events while reducing the rate of collected background events, both at Level-1 Trigger (L1T) and at High-Level Trigger (HLT).