What you're attempting is something done in editting, after both scenes in both cameras have been rendered.
What you can do, and I have done this successfully before I got an editting software, is to only use one camera that you can move from one subject to the next. What you need to do in the time line is to change the interpolation of the camera's Transition and Rotation to zero, and, voila, you got the scenes to change according to dialogue