I ask myself the same thing often, but I don't think I would migrate existing projects to JavaFX. At least not for now, and not for medium- to large-size projects. I would, however, considering JavaFX for new projects, and consider a migration again in the future and re-evaluate question based on JavaFX's progress.
At the moment, my concerns are:
Yes, we're going to 3.0 soon, but it hasn't been around this long, and still went through major changes. So for large and risk-averse corporate software, this is a relatively sore spot.
I haven't seen enough hard-data on performance differences.
Widgets and Components
I haven't seen enough gain in the new components. This can relate to immaturity, I guess. I also don't know yet how well they can be extended and composited, in contrast with Swing.
Overall, I guess hard-data on the advantages is what I'm missing to be fully convinced by JavaFX.
On the other hand, Swing is proven and tested. Yes, the API is clunky and invoking the auto-completion in your IDE on a Swing object like the JTextPane will make it weep and cry for its mommy, but, if you're knowledgeable enough, you can build awesome UIs with Swing, that are performing well (I never bought the Swing-has-bad-performance fallacy, see Romain Guy's former blog posts on Sun's blogs) and allow you to do pretty neat things.
So, before switching anything, I'd recommend you try a small prototype first, and maybe try to port some of your application's dialogs and see how it goes.