What is retrieval-augmented generation (RAG)?
Updated Oct 25, 2026
Definition
Retrieval-augmented generation (RAG) is a way for AI to answer from specific documents instead of only from what it learned in training. It first finds the passages relevant to the question, then writes the answer from those passages.
How RAG works
- Index: the documents are split into passages and stored so they can be searched by meaning.
- Retrieve: for each question, the most relevant passages are found.
- Generate: the AI writes the answer using those passages, often citing them.
Why it matters
Answers are based on your own content, so they're more accurate and can point to their source. It also means the AI can answer about things that happened after it was trained, like last week's meeting.
How Acta uses RAG
When you ask Acta a question, you choose a set of meetings, such as every call with one customer or a month of standups. Acta searches those transcripts for the relevant parts and writes the answer from what was actually said. Only your meetings are searched, only for you.
Ask questions across your meetingsFAQ
Does RAG stop AI from making things up?
It reduces it a lot, because the answer is written from real passages. It's still worth checking the source for anything important.
Is RAG the same as training the AI on my data?
No. RAG looks up your documents at the time of the question. Training changes the model itself.