Can I automatically reduce voice level peaks down to the avg level of the file itself?

My purpose is to have the voice audio file peaks closer to the same level so when I do a normalize to -3 dB those higher peaks are not drastically keeping my average level from increasing.  Right now I have been going into the individual wave form, dragging down the higher peaks toward the average peaks of the file, then normalizing.  I figure there has to be an automated way to do what I am manually doing.  I have 8-2hr length audio files that I need to do this for.  Automation would greatly reduce my time working on them.