Each of the Office applications that works with XML is targeted to a particular set of XML uses. While many people think of XML as a general-purpose format that can store any kind of information, there are some serious divisions in the way XML is used and in the practices surrounding that use. While some of these sound like the usual programming divides, where Visual Basic, C#, Perl, and Java programmers all look at the same information slightly differently, some of them are more like the divisions between people who primarily use Word to create documents and people who primarily use Access to create and present databases.
The most commonly discussed division in the XML world is the divide between documents and data. XML's predecessor, Standard Generalized Markup Language (SGML) was used primarily for document management. While having structures in documents was a key feature for organizations with huge numbers of documents like various departments of defense, the U.S. Internal Revenue Service, airplane manufacturers, and publishers, the structures were generally seen as labels applied to documents, not as structures defining the contents of the documents. Documents have to be accessible to humans as well as computers, and document structures need to be able to keep up with the many intricate structures humans create to solve particular problems.
Developers who focus on data structures typically see XML as a tool for creating labeled containers for information. While there may be some variations in that data and perhaps even some intricate data structures, the contents are generally expected to conform to the structures, not the other way around. Programmers who want to exchange data typically start by defining structures, and build code around those structures. Many program structures, especially efficient program structures, are very brittle and don't take kindly to changes because of different contexts or people adding extra layers of labels and structures.
While these two camps are often seen as separate and mutually suspicious, they can and do mix. Many documents contain some strongly structured information, like tables or lists, and sometimes data needs an escape hatch for possibilities that can't all be predicted in advance. Databases have long had fields that can support information in "rich" formats, from simple text with bold and italic to complex multimedia. XML is not a cure-all that can make all of these different views on information play nicely together, but it does offer enormous inherent flexibility for representing different kinds and styles of information. (Sadly, no XML features appear in Macintosh versions of Office.)