Everything you need to fix this is in Effects>Amplitude and Compression.

This optimises the signal as far as the level is concerned, but you are almost bound to have too much dynamic range in it. This means that to start with, you need to normalize your signal so that the peaks are almost at 0dB. So what you have to do is optimise your speech recording to fit in with these essentially restricted parameters. So that gives you a signal to noise ratio at the very best of about 48dB, and that's worse than an old cassette recorder! You have to bear in mind that 8-bit audio has a significantly higher noise floor than 16-bit audio - if you want to work out where it is, it's 6.02dB/bit.

Nothing really - it's what you haven't done that makes the difference here. The original file was a 44100 Hz 16-bit Mono that was crystal clear. These settings worked for my Verizon VOIP system, but the quality is horrible with a lot of hiss in the background.