RAG questions, answered
Short answers to the questions we hear most often about answering questions from company documents.
Do we need a vector database?
Not necessarily. OpenSearch, Elasticsearch and PostgreSQL with pgvector all support vector search. Start with what you run today; move to a dedicated vector database only if scale or features demand it.
How accurate will the answers be?
That depends mostly on document quality and retrieval, and it should be measured, not promised. With good documents and a tuned system, most answers to questions covered by the documents should be correct, and the rest should say they do not know.
Can it answer in Hindi?
Yes, with a multilingual embedding model and an LLM that handles Hindi well. Test with real questions from your users.
Will it show where answers come from?
It should. Every answer should cite the documents and sections it used, so users can check.
What does it cost to run?
Mostly the LLM usage and the search infrastructure. For a few thousand employees, running costs are usually modest compared with the time saved. The bigger cost is the initial work on documents and testing.
Can RAG take actions, like raising a ticket?
That is the next step: an agent that uses tools. See agentic AI.
More RAG guides: How RAG is built · Preparing documents · Better retrieval · Permissions · Measuring answers · RAG FAQ · Use case: HR policy assistant
Want answers from your own documents?
Tell us where the documents live, who should be able to ask, and ten questions people ask today. We will come back with a plain view of what it takes to answer them well, and safely.