I operate on a similar line but have a cascade of folders

client -> matter -> subfolders.

this means that i never get a list of active files from my folder architecture but I use other tools for that (more on which another time).

also like you i rely mainly on unstructured search for retrieval of files. I recall attending an in-house meeting at my alma mater (magic circle firm in the uk) where I had been invited to advise on how to implement structured search for their new infobank. This was in 1999. a lot of thought had been put into it and the workgroup were clearly very pleased with what they had done and wanted a pat on the head. They were a bit put out when i told them that my advice was to ditch it all, ditch the parsing and tagging plans and instead OCR every document and buy a fulcrum server for FTI and natural language search. Nowadays I would firmly add tagging back in, at two levels: a user can add their own tags to material and if two or more users add the same or similar tag, it would essentially become a ‘public tag’.

Turns out they implemented their idea. then 18 months later, ditched it and implemented mine.