RAG vs. Agentic RAG
NewBoth answer questions from a company's own documents, and many resumes now call any RAG system 'agentic'. Plain RAG runs one search and passes the results to the model. Agentic RAG lets the model decide what to search for, search again, use other tools and check its answer before replying. It handles harder questions but costs more, runs slower and is harder to test. It is also newer, so long claimed experience deserves a closer look.
How to tell them apart on a resume
RAG
Embeddings, a vector database, chunking, a retriever, top-k search, a prompt with retrieved context, a chatbot over documents.
Agentic RAG
Query planning, multi-step retrieval, tool calls, LangGraph or other agent frameworks, self-checking, routing between sources, evals for each step.
The question that settles it
“Did your system run one search and answer, or did the model decide what to search for, search again and check its answer before replying?”
Each term in plain English
- RAG
A technique that lets an LLM pull in relevant information from an external source (like a company's documents) at the moment it answers, instead of relying only on what it memorized during training.
Ask “Walk me through how you chunked and retrieved data for a RAG pipeline you built.”
- Agentic RAG
RAG where an AI agent decides how to look things up instead of running one fixed search: it can rewrite the question, search several sources or tools, check whether the results answer it, and search again before replying.
Ask “Why did your RAG system need an agent rather than a single search, and how did you show the answers actually got better?”
- AI Agent
An AI system that can plan, take multiple actions (like calling tools or APIs), and adjust based on results, rather than just answering a single prompt.
Ask “What made the thing you built an 'agent' rather than a single prompt-and-response call?”
- Vector Database
A database optimized for storing and searching 'embeddings' — numerical representations of meaning — so a system can find content that's semantically similar, not just keyword-matching.
Ask “What made you choose a vector database over adding vector search to a database you already had?”
Open the full tool for the other look-alike pairs, role profiles, and the JD decoder.