In addition to a minimum sampling rate of 8 kHz, because of the dynamic range associated with speech, you will require more than the 8 bits of resolution available with the GP32, for reasonable intelligibility. Resolution of 12 bits would be optimum. Special codec devices, as used within the telephone system, effectively provide a logarithmic compression of the signal down to 8 bits per sample, and provide for expansion when the sample is decoded back to analog.
I suspect you would need to use an external codec device for each of the eight channels, but this is probably outside the scope of your current project.
The ADC on the HC08s use multiplexed inputs. That's how they have multiple ADC channels with only one (sometimes 2) hardware ADCs. You cannot however, sample channels simultaneously with a single hardware ADC.
You set up the ADC, set the input channel, read the data, then move to the next channel.
I wouldn't go below 8Khz sampling rate for voice, so make sure you can get a sample from each channel fast enough.
Why ASM? CW 5.1 has a free c compiler and using the processor expert tool, it can set up and write the ADC sampling code for you.
This can be done on an 8-bit uC, however your application sounds like a perfect fit for a DSP. ADC sampling, muxing an output stream and demuxing an input stream can be done in hardware. That's a perfect app for them.
I'm sure if you check out Freescale's 56F8xx family, you can even find code samples similar to what you want to do.