Quicksilva: What I do is pretty simple. In whatever I draw which I will then fully shade or color (meaning I don't do this for comic art, which is very high contrast and has different inking and shading tropes most of the time) after I have my general shapes and lineart pretty much there I faux-3d render it by hand. Meaning I actually connect tris from point to point and make the geometry more complex step by step.
I mocked something up very fast because I should actually be doing comics instead of tutorials, but a few images are worth more than just dry text. Let's take the below doodle:
Then I start doing this, just more detailed. Once you're used to it a bit it's pretty mindless. I might not go to as much detail for simpler bits where I know how I'll behave. The key point in all of this isn't to just do this every time, it's to *think* it even when you're not doing it. Abstraction of something we're used to think as a bunch of symbols to pure geometric planes. Less polygons are better than more polygons.
Then select just a few shades and group polygons that, according to the lightsource would be closer to the same shade than the next jump towards darker or brighter. This is pretty basic, they teach this at any fine art course (of course not this computery and much more proficiently, but I learned by myself pretty much so)
I won't keep going for the whole thing, just some explanations of how I handle basic volumes. Once all value range has been applied then according to whatever style I'm going for I might blend between triangles or I might not. In pixel art that's usually where I start dithering between clusters.
This is the later without the lineart. Obviously there needs to be more of a bridge between the higher values and the middle values, like 2-3 colors but I don't have the time to do this, I hope the concept is clear enough as it is.