I personally vote against the use of Story points.
For the time being, I have my team using canned option time estimates (15m, 30m, 1h, 2h, 4h, 1d) anything larger gets broken down until it fits.
I don’t think that Story points would make estimation more accurate.
I don’t like that a Story point in Project A is not equal to a Story point in Project B.
Lastly, while having a planning poker session around User Stories is feasible, what do we do with bugs as they are reported? All bugs are not made equal; many are trivial to resolve, some (in theory) could require redoing a whole Story worth of work.
In summary, I don’t see what they add, but I see what they subtract.