# How does agentic RAG differ from basic RAG?

Source: https://zilliz.com/ai-faq/how-does-agentic-rag-differ-from-basic-rag
Author: zilliz

Basic RAG retrieves documents first, then generates answers; agentic RAG lets the agent decide whether to retrieve at all.

Key differences:

| Aspect | Basic RAG | Agentic RAG |
|--------|-----------|-------------|
| Retrieval trigger | Automatic for every query | Agent-driven, context-dependent |
| Query handling | Fixed retrieval method | Rewrite, decompose, adapt queries |
| Iteration | Single pass | Multiple retrieval loops |
| Relevance evaluation | User evaluates output | Agent evaluates and re-retrieves |
| Error handling | No recovery from bad retrieval | Agent detects and corrects |

Agentic RAG excels when:
- Queries require multi-step reasoning ("What products did we sell to this customer, and compare their margins to industry averages?")
- Documents may be irrelevant or conflicting
- Complex workflows need intelligence about when external knowledge is needed

For production agentic workflows, the vector database becomes the agent's dynamic memory—not a static document store. This is the 2026 enterprise baseline.

Zilliz Cloud handles this seamlessly: managed infrastructure, automatic scaling, and enterprise-grade reliability for agentic workflows of any size.

**Related Resources:**
- [What is Zilliz Cloud](https://zilliz.com/what-is-zilliz-cloud)
- [Agentic RAG with Claude and LlamaIndex](https://zilliz.com/blog/agentic-rag-using-claude-3.5-sonnet-llamaindex-and-milvus)
- [Local agentic RAG with LangGraph](https://zilliz.com/blog/local-agentic-rag-with-langraph-and-llama3)
- [Start free on Zilliz Cloud](https://cloud.zilliz.com/signup)