How do you approach software modeling?

I'm trying to build a good overview of a medium-complexity distributed system which is currently, documentation-wise, a bit of an anarchistic blob of individual text documents, code comments, and a whole-lot of databases stored individually into each developers brainDB. What is todays gold standard of doing software modeling? The first reaction is a flashback to CS classes and digging up an UML tool, but interested to hear if things have progressed in this area..