This is probably more appropriate for the biology stackexchange.
"Sensitive" and "insensitive" are dose-dependent descriptors and the relevant dose ranges will depend on the off-target effects (assuming they're even known). If you just have to randomly pick some thresholds, then an IC50 in the pM to nM range would tend to indicate a sensitive line, while IC50 in the µM or higher range is probably insensitive. The reason for this is simply that off-target effects increase with dose. Note that these vague ranges are just ballpark estimates, there necessarily exists no strict definition.
Realistically, if you have a bunch of cell lines and some are likely sensitive to a compound and others not, then simply graphing the IC50 values will likely show you the group structure and whether there's a reasonable discrimination threshold. There are statistical techniques that can be applied here, but they're overkill and you should look at a graph anyway.
If you're not overly familiar with the compounds and/or biological system that you're using then just sit down with a biologist/chemist familiar with the specific project and show him/her the IC50 values. He/she can then give you some insight into reasonable ranges.