Semantic Search
Semantic search matches on meaning rather than wording, so a document can come back as relevant without sharing a single word with the question.
Keyword search is precise and literal: it finds what you spelled. That fails constantly inside a company, where the same thing is an SOW to one team, a scope doc to another and "the Acme agreement" in the thread where it was actually decided. Semantic search compares embeddings instead of strings, so the wording stops mattering.
It is not a replacement for keyword matching, and production systems generally run both. Semantic retrieval is weakest exactly where literal matching is strongest — error codes, ticket numbers, product SKUs, anything where the exact string is the point. The useful system is the one that can do both and choose.