ok.
lets start:
we need a 3d model of a phone, lets say, an iphone (properly modeled, textured), then we need a shot of a moving finger which we will use later on. lets say we have everything.
after we have the idea, what we should do with the finger(adequately having this shots on green screen so we can remove the background) we animate the phone in a 3d package, after we are satisfied with the result, we render the phone sequence out from the package (properly lighten and rendered with alpha channel - ready for compositing)
Open after effects or any other compositing software, we put the a white background as solid, then we put the phone sequence on top of the background, the sequence goes, we put the finger where we want him to interact with the phone, since our phone has no animations on it, we can put them in 3d or we can add them in after effects, a little bit of tweaking and there you are, you got your 3d rotating phone with a finger and a hand interacting with it.
if we want to comp anything with the phone in aftereffects, we need the rotation and camera information from the 3d package, lets say cinema4d, we can export ae proj file out, with all the info wee need, and later use that data in after effects to have the perfect sync. Other way around is to have a properly UV for the phone, and put in animation sequence on the screen material, the render it out, then we have just to sync the figer with the animation.
hope this reveals some magic ;)