In this InfoWorld article, Jon Udell gives the 10 things you should know about InfoPath (ne'e XDocs). There are a couple of points I thought deserved emphasis:

• The product includes a full-blown DOM, not just a SAX API which means that you should be able to manipulate the XML, not just read it, from an outside program.
• There's a visual XSLT tool. As Jon points out, XSLT is powerful, but difficult to use (unless you're a Prolog programmer---then its old hat).
• InfoPath can generate a schema from an XML snippet. You may not want to use this generated schema as your final version, but its a useful way to quickly get something that's pretty good and then refine it later.

Jon claims that this is a pradigm shift and I agree. Having a tool that stores unstructured data in a semi-structured, common format and is likely to be widely used because of the links to the new Office suite is a powerful combination. Moving the vast quantities of unstructured data to an easily accessible semi-structured format will bring huge changes to the enterprise. One example is the ease with which you can replace what is now a custom application. There are companies waiting to be born to take advantage of this change.