Khovanov and Seidel defined an action of the braid group by autoequivalences of a certain category of projective modules over the so-called zigzag algebra. Taking the Grothendieck group, one recovers the famous Burau representation, but unlike the latter, Khovanov-Seidel representation is faithful. In work with Licata, I showed how to use Khovanov-Seidel representation to extract metric data on braids. Building upon this idea, I'll try to convince the audience that such categorical tools should play in the larger context of geometric group theory.