Instead of tagging (which takes effort) you could also use a good search algorithm.
Probably depends on your use case, but my experience is that getting good results with text search often takes a lot more effort than tagging/classifying the documents when you get them would. It takes effort to craft a good query. Terminology used varies. "Oh, Sally uses a different term than I do, what was that?" If you tag or put the files into folders then you can standardize.
I worked as a patent examiner in the past and using only text search for that would be considered negligent there as it would miss a lot of documents. I ultimately used a combination of text, classification, citation, and AI similarity search techniques. Each has strengths and weaknesses so using all of them makes sense.