A concrete example
Classic RAG searches once and answers. Agentic RAG searches, evaluates, and searches again with different keywords if the first pass isn't enough.
Why it matters
On complex questions (multi-step, multi-source), agentic RAG often doubles the quality of classic RAG.
You'll see it in research assistants, Azure agentic retrieval, LangChain, and watch/curation agents.
Don't mix it up with
RAG: RAG lets an AI answer using documents pulled up at the moment you ask the question.
Agentic workflow: An agentic workflow is a sequence of steps where the AI plans, uses tools, checks its work, and keeps going until it hits the result.
Common mistakes
- Running agentic RAG on a simple question (slow and pointless).
- Not capping the number of iterations.
- Confusing agentic RAG with tool use (close, but not the same).
Quick checklist
- First I check whether the word names a concept, a tool, a risk, or a metric.
- I tie it to a concrete case: Classic RAG searches once and answers. Agentic RAG searches, evaluates, and searches again with different keywords if the first pass isn't enough.
- I keep the main trap in mind: Running agentic RAG on a simple question (slow and pointless).
Quick questions
What is Agentic RAG in AI?
Agentic RAG lets the agent decide when to search, what to search for, and when to search again.
Where will I run into Agentic RAG?
You'll see it in research assistants, Azure agentic retrieval, LangChain, and watch/curation agents.
Which word should I read next?
Start with RAG, Agentic workflow, Retriever.