The recording is perfect except everyone has to max their volume to hear it. Bring it up properly — measured, clip-safe, and without the pumping that "auto-volume" apps add.
Speech mode measures the loudness of the parts where someone is actually talking (quiet passages don't drag the average down) and raises the whole file to a standard speech level. If that gain would push any peak past −1 dB, a soft limiter catches just those peaks instead of letting them fold into distortion — you'll see a note below the result when it had to. Best order of operations: denoise first, normalize second — otherwise you're amplifying the hiss along with the voice.