Are you just using Newton's law of gravity? If so, what numbers are you using? With 10
12 solar masses for Andromeda and a megaparsec separation, I'm getting it to be about a factor of a billion less:
a \sim 10^{-13}~m/s^2
My number is more like what I would expect, since accelerating at that rate for the age of the universe produces a velocity:
v \sim at_u \sim (10^{-13}~m/s^2)(10^{10}~years) \sim 100~km/s
which is of the same order of magnitude as the current velocity between the galaxies.
Also, I think it's worth mentioning that the masses of the Milky Way and Andromeda galaxies are still very uncertain, so estimates of the acceleration would be as well.
BTW: The unit calculator in my signature makes these calculations a lot easier.