Use microdata to mark up useful data
It makes sense to use schema.org markup if you are representing any of the supported data types, and when that data is likely to be useful.
Be aware that microdata makes file sizes bigger
How you evaluate whether data is 'useful' or not is a matter of personal taste. Just be aware that microdata can add weight to a page, so using it for everything isn't necessarily a great idea.
To give you an idea of how much code it can add to a page, here's an example from the ImageObject spec:
Without schema.org microdata:
<h2>Beach in Mexico</h2>
<img src="mexico-beach.jpg" />
By Jane Doe
Photographed in Puerto Vallarta, Mexico
Date uploaded: Jan 25, 2008
I took this picture while on vacation last year.
With schema.org microdata:
<div itemscope itemtype="schema.org/ImageObject">
<h2 itemprop="name">Beach in Mexico</h2>
<img src="mexico-beach.jpg" itemprop="contentURL" />
By <span itemprop="author">Jane Doe</span>
<span itemprop="contentLocation">Puerto Vallarta, Mexico</span>
<meta itemprop="datePublished" content="2008-01-25">Jan 25, 2008
<span itemprop="description">I took this picture while on vacation last year.</span>
The extra code isn't necessarily a bad thing, because humans aren't likely to have to read it, but it's worth considering the effect of extra microdata on page load time over mobile networks, for example. For this reason, it might be best to consider what data on your page is likely to be useful to automated scripts and machines, rather than adopting an all-or-nothing approach.
Google recommends you use microdata now
Google already uses schema.org microdata in certain search results, and they say that the number of results (and companies) using it is only likely to increase. From their schema.org FAQ:
Search engines are using on-page markup in a variety of ways—for
example, Google uses it to create rich snippets in search results. Not
every type of information in schema.org will be surfaced in search
results but over time you can expect that more data will be used in
more ways. In addition, since the markup is publicly accessible from
your web pages, other organizations may find interesting new ways to
make use of it as well.
They also point out that they see schema.org microdata as the future of structured data markup, and that they don't plan to support 'competing' formats such as RDFa and microformats:
Historically, we’ve supported three different standards for structured
data markup: microdata, microformats, and RDFa. Instead of having
webmasters decide between competing formats, we’ve decided to focus on
just one format for schema.org. In addition, a single format will
improve consistency across search engines relying on the data. There
are arguments to be made for preferring any of the existing standards,
but we’ve found that microdata strikes a balance between the
extensibility of RDFa and the simplicity of microformats, so this is
the format that we’ve gone with.
Browser makers are gradually supporting microdata
Browsers are slowly supporting microdata too, so it's likely that we'll start to see better integration in software, which might one day help to automate the process of adding a contact to your address book straight from a web page, or adding an event to your calendar, for example. (Support in the Firefox and Opera browsers is in progress.)
Consider RDFa Lite
Google gave advance notice of support for RDFa 1.1 Lite in November 2011. Although RDFa Lite isn't as descriptive as schema.org microdata, it's easier to learn and implement, and might prove more publisher-friendly for users wanting to adopt microdata fairly quickly. Details of the proposed spec are on the RDFa Lite draft page.