Assuming your company is large enough that people don't sit with each other, there often comes the painful realization that Sue has just spent two months on a project, only to discover that Pedro has a deep interest in the topic and could have helped Sue shave off a couple weeks' effort. Expert location services come out of stories like this. And they have been considered a staple of knowledge management for quite a while.
From what I have seen, expertise location started with company yellow pages that were converted to electronic use on an internal website somewhere. The variety of information that could be housed in such a system has exploded, and there are a resulting multitude of mechanism for building and maintaining them today.
In general, expertise location system make a connection between "what" and "who," so that searches for "what" turn up the people who might know something. Search for Subject X, and out pop names and links to fuller profiles of those people. The systems I find more interesting have this expertise information coordinated with other searches, so that searching on Subject X turns up content on that topic, but also relevant people who know something about that content (and why they might know it).
There are two basic techniques and a third hybrid mechanism: Manual and Automated. which is almost too obvious to explain. In manual systems, all the information is entered and created by hand - most often by the "expert" or possibly a manager. This information can be the basic name, rank and serial number (hopefully pulled from the company directory); resume; a ranking within a fixed taxonomy of skills; free-form textual information; etc. The obvious drawback of manual systems is the requirement to keep them up to date and provide some level of verification that people are who they say they are.
Automated systems attempt to generate the expertise information, based on information that is already available: email, documents, browsing history, etc. The difference here is that with the wealth of information available, these systems have the ability to build a lot of connections that aren't possible to do in a manual system. The question is, which of these or how much of this do you include in an automated mechanism for establishing credible expertise? And, once you decide, can you get over the "big brother" factor, even within the organization that owns the work content. One of my biggest concerns with these systems, is that if they don't see everything people do, they have the potential to skew the information to only what is housed in the repositories visited.
Hybrid mechanisms essentially give users the first right of review of the profiles that the automated systems build. In most cases, the automated system goes and does its thing, and the users can tweak how their expertise is presented. This particularly makes sense if the system provides a list of "expert topics" assigned to the user: they can simply indicate their disagreement with the categorization. A somewhat more active system might allow a user to indicate which content is most relevant to their areas of expertise. (I just read about this particular version on Technology Review's comments from the CHI2009, A Smarter Way to Dig up Experts.)
The Question. If you are going down the path of expertise locators, be sure to check with yourself about why you need such a thing in your organization. Are people having trouble finding people that must be working in the company? Are the local experts unwilling or unable to provide the right level of assistance? Are people spending too much time recreating solutions or answers that already exist? These are all symptoms: have you established an underlying reason for these observations or behaviors.
And if you decide that it's the right thing to do, some other things to check: Is the corporate culture such that people can ask for help and receive it? Is professional expertise guarded heavily because that is how people are measured and rewarded? Do you expect everyone to be listed as an expert? Do you expect everyone to be available to respond to requests? What happens when they don't (for a myriad of reasons)?