I have the following scenario: the user picks a set of entities from a palette (each palette entry can belong to a different class, and each class can belong to a different subclass).
After extracting the palette entries he is interested in (for example, by drag&dropping them from a tab widget divided by classes to a "receiver" list), he will press a button or something that will activate a computational machinery working on the provided elements in the list. This machinery will produce a list of results, and each result must be selectable and displayed in more detail. The current naive idea I have is the following:

I am not particularly excited about it. Actually, I think it's horrible. The workflow starts from the bottom left, moves in the upper left, then goes to the bottom right, and the specific entry is displayed in the top right. Rearranging them may be an option, but then I may encounter both problems with real-estate, and with the fact that the focus of the operation is the final result.
Another option would be a wizard, but a wizard generally is aimed at performing a step-by-step operation operation once. In this case, the user is free to delete the generated output and experiment with a different selection out of the palette, then check the new results.
How would you design such UI to be more appropriate for the workflow I presented ?

