SAS means "stability augmentation system". They are designed to reduce angular rates (pitch/roll/yaw rate). This makes the helicopter less likely to move quickly from a desired flight condition. SAS doesn't account for control cross coupling such as yaw response due to collective input.
Sometimes you will see control mixing where a collective input will add approximately the right amount of pedal. Pilot input is still required, but smaller inputs.
To create a system that produces no yaw with collective input requires quite a bit more effort and expense. The helicopter needs to maintain a heading, so a reliable heading source is required (gyro=more expense). The inputs to the control system would be larger than a SAS would require, so now you need control actuators that have large authority (actuators=more expense). The logic of the system still needs to know when you do want to change the heading, so you aren't fighting the AFCS whenever you try to fly yourself (computer=more expense). There's more to it, but I think that covers the big points.
Obviously not impossible, and most medium or bigger helicopters built today do this and so much more and do it very well. However, they are expensive. As Rinker said, pilots quickly learn to compensate for this. So all you really need is a good excuse to justify the cost.