I had another opportunity to spend some time with Cogito. Cogito's current products represent a form of content management for engineering documents (everything from schematics to detail drawings). As we were talking, I began to see them as doing knowledge management and collaboration, but from the inside out.
When a traditional collaboration company (maybe too young a field to be calling anyone "traditional" but bear with me) like Groove approaches collaboration and knowledge management, they view the archive of team data as the artifact and build a meta model of the archive in an attempt to provide an understanding of the information to the team, allow them to make better use of it, and provide a collaboration vehicle. In many cases, that archive of information is a collection of representations of some mental concept that is collectively shared by the team.
As an example, consider a Boeing 777. There are millions of engineering drawings and other documents and hundreds of databases that collectively represent the archive of information about any particular 777 design. The design process consists of engineers and others building this archive of information to represent the mental map they collectively share of what a 777 is. Now, suppose that you could build a representation of that mental map from this archive in such a way that the entire archive could be thrown out because any piece of it can be regenerated at will. What's more, any changes to the artifacts in the archive update the model and consequently any other document associated with that artifact is automatically updated as well.
This represents an alternative approach to knowledge management that I think of as "inside-out." Instead of building a model of the archive on the outside, you build a model of the concepts---the things inside the artifacts that make up the archive.
This may all sound too good to be true, but Cogito has real contracts with real companies producing these models on the scale I've described here. Their revenues are modest to date, but have a significant partner that should provide an excellent channel for selling their technology. I think that deep down, there's a connection between what Cogito is doing and what the semantic web is attempting, but two and one half hours wasn't enough time to ferret that out.