I'd recommend starting small, as the learning curve for WPF/Xaml is pretty steep, I reckon a competant C#/WinForms developer will take 6 months to get comfortable - so your "...big application and lots of business logic..." is probably not a great candidate to start with.
OTOH if you're team is already upto speed with WPF, the real first step is deciding what features the users want that can be best served by the graphical niceties of WPF. If they really do want 3D spinning forms, then implement a good working prototype of that, if they want resizing documents, then do that. This is an elephant of a problem, you can only solve it a small piece at a time.
A key thing to decide really early on is whether to go MVC or one it's various offspring. That decision will shape where you put your code. As WPF is heavily dependant upon binding, you can't move forward until you know how/when/where that binding is going to happen.