Choose how hard to clean files for AI

An assistant can only cite what the files actually say. Match the cleaning effort to the size of the pile.

Good

A short pile you can clean in chat. Slice long files, tighten each piece, and merge them into one source.

Better

A large library. Use a pipeline so you are not slicing every section by hand.

Best

A shared knowledge base. Structure the files so answers stay tied to the right policy, role, and source.