A de-esser does exactly what the name implies: it attempts to remove sibilant frequencies from your vocal syllables: the extended trails of “sssss” sounds, etc. So the short answer is if you hear excess sibilance in the vocals, you might want to use a de-esser.
However it’s good recording technique to try to reduce sibilance at the source first. Your vocalist can learn singing techniques to reduce sibilance. But even with solid techniques, it’s possible that EQing and compression can end up bringing up the sibilance in a track, and a de-esser in the vocal processing chain will be a good idea.