I haven't covered this! Here's how it works:
The side chained input is actually an envelope follower where the amplitude of the input is immediately transformed into a control. You can map that control via the EXS24 router to several different targets... one of which is pitch. That "pitch" is not able to be mapped to exact pitches. However, you can route the side chained control to sample select where you can probably get it to play a custom sample... from the manual:
Sample Select
Modulates the sample (zone) that is played.
By default, Sample Select is controlled by velocity through the default Velocity to Sample Select modulation routing. This means that the received note velocity value determines which of the layered zones (in different velocity ranges) is heard as you play the keyboard softer or harder.
You are not limited to using velocity, however, to determine which sample is played. You could assign the modulation wheel source to the Sample Select destination, or use both velocity and the modulation wheel.
If you choose a continuous controller such as the modulation wheel, you can step through the velocity layers during playback. In this case, make use of the crossfade (Xfade) parameters to create smooth transitions between velocity split points.
When using multiple modulation sources, be aware that these can cause all velocity layers to run simultaneously using up as many voices as there are layered zones. The CPU usage will increase accordingly.
Try it!