In general you would choose Part workbench, when you want to create some solids, use Boolean operations in between them, attach some sketches and being able to externally link to everything from Sketch, hierarchy in between added features mostly being horizontal. If you want to work with a geometry, such as point, line, face, there is Part Shapebuilder
, hopefully in the future there will be a parametric counterpart. In addition there are some good commands in Part workbench, like Part Defeaturing
In general you would choose Part Design workbench, when you want to create a single solid and the steps needed to create it involve adding a plethora of (interconnected) features. Added features can have more vertical hierarchy, but that is not a requirement, as you can use datum features as a common reference.
In the end you can achieve most of the tasks in both workbenches, and can use both in an unconventional way. Therefore people usually just choose the workflow they prefer and the workbench it provides it, for being able to complete some specific task. Lets say you are in a room full of cartoon boxes. You want to just take some boxes and to manually arrange them to represent some structure, maybe to take scissors and cut a part of one box away. Or you would like to take a pencil and draw some lines in between them. For that best to start in Part workbench. Lets say you want to create a cartoon box. For that best to start in the Part Design workbench.
Can't we just merge them in the future? For sure it could be done, but the merged workbench needs to support both workflows, horizontal and vertical one, should be able to handle workflow including geometry such as points, lines, faces and solids. Based on the past discussions and evolutionary steps it is not likely this will happen anytime soon. What likely will happen is gradually the difference in between this two workbenches will get blurred again a bit. As over time both will likely adapt to paradigms people would like to adapt to.