The Knowledge Layer
for AI Agents
One unified endpoint. Multi-source retrieval. Permission-aware grounding. Foundry IQ turns Azure AI Search into an agentic retrieval engine that plans, searches, and synthesizes — so your agents don't hallucinate.
Retrieval as a reasoning task.
Traditional RAG sends one query to one index. Foundry IQ treats retrieval as a multi-step reasoning process — an LLM plans sub-queries, executes hybrid search, self-assesses quality, and synthesizes grounded answers with citations.
The LLM analyzes your query and conversation history, then decomposes complex questions into focused sub-queries. Each sub-query targets specific knowledge sources, enabling multi-hop reasoning — answering questions that require connecting information across multiple documents.
Three pillars of intelligent retrieval.
Knowledge Bases, Agentic Retrieval, and Grounded Answers — replacing fragmented RAG pipelines with a single, reusable knowledge layer.
Topic-centric collections connecting Blob Storage, SharePoint, OneLake, and the web into one reusable endpoint. Automatic indexing, chunking, and vectorization included.
Microsoft Tech Community
Self-reflective query engine powered by LLM reasoning. Plans sub-queries, selects sources, runs parallel hybrid search, and iterates until quality thresholds are met.
Microsoft Tech Community
Every answer includes inline citations traced to source documents. Agents reason over raw data — no hallucination, full auditability.
Microsoft Learn
Enterprise AI answers three questions.
“Who talked to whom about this yesterday?”
Tracks employee activity across Teams, Email, and Graph.
“Where is that data?”
Locates and maps data across Data Lake, OneLake, and enterprise sources.
Phase 3“What do the documents say about this?”
Searches KB with agentic retrieval — plans, retrieves, assesses, synthesizes with citations.
★ THIS DEMOThis demo deep-dives into the third question — how Foundry IQ searches your knowledge base and synthesizes grounded answers.
Fabric IQ locates enterprise data → Foundry IQ searches it for answers. Two questions, one pipeline.
See it in action.
From Knowledge Base query to Trace visualization — experience the full pipeline.
Launch Demo Suite