I like the idea, but right now I don't really see the benefit of AIR. It's another runtime to download and install before being able to use an application, that just adds one more step. The only benefit over a browser-based web app is desktop integration such as the local file system, clipboard access, and some persistent local storage. Then you lose the benefits of web apps: access from any computer with a browser, all data stored remotely, no need to install anything, etc. That's just for the users.
For developers, it creates a whole new set of problems. They have to maintain, test, and deploy two apps with different codebases. They have to deal with users having different versions of the runtime and different versions of the app. Seems like a lot of overhead for not much in return. So I think the use case for AIR will primarily be for just creating small, helper apps that are auxiliary to a web app. I can see the utility in an file uploader for Flickr or something to that effect. I guess we'll see what happens in the next few months and what apps come out. Maybe an AIR app will be released that will really blow me away and I'll see the point.