Your friend is confusing sampling with encoding. Your sampling rate is determined by the frequency content of the signal being sampled. the bit depth, ideally is limited by maximum signal excursion on the high end and by noise on the low end.
With digitization you have distinct code encoded in binary format for each signal level. If you have a discreet set of steps of say 0, 1 ,2, 3 ... they are represented 0 = 000, 1 = 001, 2= 010, 3 = 011 etc. in the digital domain. But the 0,1,2,3 steps can easily be 0, 0.1, 0.2, in some signal quantity, like voltage. This is dictated by the conversion constant of the convertor, which might say 100 mv/DN (millivolt per Digital Number).
If you are fitting an 8 bit sample in say 2 volts of signal you have 256 states in 3.0 V and thus 7.8125 mV/DN.
In fact you can have a sampled system that is still analog. These devices (uncommon now) are called BBD (Bucket Brigade devices) AKA CCD (Charge Coupled Devices) - which found their main use as Imaging devices, but started out as signal sampling and processing devices.