I believe strongly in the model of WebApps, but we have to look at the web for it's real productivity benefits, connectivity and taping the information available. The apps that will take hold on the web are the ones which needs a certain level of connectivity, ie they make use of the fact they are connected in a significant way. For example - Reviewing files is very important, productivity when on the road is reviewing, not producing as such, we should have services for reviewing word/excel files safely in a public environment, it would beat doing this on a blackberry or treo.
AjaxWrite is going to improve, but when are you going to write a document of more than a few paragraphs in a browser window? Maybe when you write an e-mail, so yes I could see Outlook going web only. Filling in a form is another place where you spend decent amount of time writing stuff in a browser window. Another one document reviews, quick and dirty comments or edits to fire back to the author of the document. The competition in those cases is not Word or Excel, they are not the targets, or they are the wrong targets in my mind, but They are easy targets, they are Microsoft, and they get you mileage. Innovator's solution as a good case study on what the blackberry really competes with, I think it applies in this case as well.
I think Ajax applications will develop into a new category, a good interface for manipulating, editing, creating information of and on the web. There are sources of informations organized and not organized on the web that are just asking to be put to use. Ajax apps will leverage webservices in new ways. This is both a larger and narrower mission than creating documents. The productivity apps are a few the basic elements from which they will feed. They will take over from where the productivity apps stop, and go in new ways. Trying to replace instead of complementing them may be the only way to understand where the elements all fit in the picture unfortunately.