Make the full-volume speakerphone sound clear, without the "clipping" of voices. It can be easily be achieved thru digital processing, so that when you put your speaker-phone to full volume, it will actually sound good, and still be useful against the surrounding noise.

Technical notes: It could be accomplished by adding software-only or a software/hardware component to cellular phones. The algorithm should be one that maximizes the energy but keeps the phonetics. So sounds like S and V will be simply magnified, but sounds like M and R would have to be modified so that you still percieve the original phoneme but never have the sound clipped.