As computing power gets better, I believe that tools can start helping us make better time-critical decisions, instead of simply taking an input and executing.
For dense and complex geometry, determining where to place each joint can be a difficult and tedious task. Most automatic skeletonizers don't do a good job of determining the proper placement. What if a computer could label the different parts of the mesh (like body parts) and generate a skeleton from that?
I'm currently in the process of implementing Jyh-Ming Lien and Nancy Amato's recent research in shape decomposition. You can see my progress, as my plugin identifies the critical points where body parts are likely to be separated. This certainly isn't a perfect solution for rigging, but it can accelerate the process for complex and tedious models.
The mathematical properties of geometry are somewhat unexplored in computer animation, and they can be used to accelerate the rigging process. Convex hulls are a useful tool for extracting some of those properties, and I'm using them for my automatic skeletonization plugins.
There are a few convex hull generators available for Maya, but they're susceptible to problems with nearly planar faces, such as those on the sides of buildings. My implementation is based off of John Lloyd's QuickHull3D, which accounts for the imprecision of floating point numbers.