Dense retrieval uses vector embeddings to find semantically similar information rather than matching keywords. The system converts both the query and stored documents into vector representations, then finds the stored vectors most similar to the query vector.
Dense retrieval enables retrieval based on meaning, so a query about "overdue payments" can surface documents about "accounts receivable aging" even without shared keywords. This is what makes it so valuable in enterprise knowledge environments where different teams and documents use different terminology for the same concepts.
