I'd think the trigonometry actions could work here- if you imagine the ball's motion as a vector, and you want to apply a force in that direction, you should be able to compute how much force in the world x,z axes you need to apply using trig. Away from computer right now so can't advise on the correct action, but I've done this myself in the past. The advantage here is if the ball changes direction as force is applied, it will still apply with respect to the new rotation.