Foundry IQ in Copilot Studio: Connect Your Enterprise Data

Introduction: AI agent intelligence depends on the quality of accessible data

IT teams deploying Microsoft Copilot Studio agents consistently face the same challenge: how do you connect an AI agent to organizational knowledge without compromising security, compliance, or response quality? Microsoft addresses this challenge with the native integration of Foundry IQ (formerly Azure AI Search Knowledge Bases) directly into Copilot Studio.

This new integration allows developers to select an existing Foundry IQ Knowledge Base as a knowledge source for their Copilot agents — all with governance and permissions already built in, requiring no additional user-side configuration.

Product Context

Foundry IQ is the knowledge management component of the Microsoft AI Foundry platform, built on Azure AI Search. It acts as a unified intelligence layer between your data sources and your AI agents. To learn more about retrieval capabilities, see the official Foundry IQ documentation.

What Foundry IQ brings to your Copilot Studio agents

An enterprise-ready knowledge layer

Deploying AI agents in production in an enterprise context requires much more than a simple data connection. Foundry IQ was designed specifically to meet the requirements of organizations subject to strict regulatory constraints. Here are the security and governance capabilities built in natively:

Customer-Managed Keys (CMK): data encryption with your own keys, hosted in Azure Key Vault
Access Control Lists (ACL): granular access control at the document and content segment level
Network isolation: network isolation via Private Endpoints and Virtual Network Service Endpoints
Microsoft Entra ID: centralized authentication and authorization via Microsoft 365 identity
Regulatory compliance: FedRAMP High, SOC 2 Type II, ISO 27001, and other enterprise standard certifications

For a complete view of Azure AI Search (Foundry IQ) security capabilities, see the official security documentation.

Advanced retrieval capabilities for more relevant answers

Foundry IQ integration goes beyond security. It also brings information retrieval mechanisms that far exceed traditional RAG approaches:

Agentic retrieval: automatic query planning with parallel federation across multiple knowledge sources simultaneously
Iterative query planning: decomposition and iterative refinement of complex queries to maximize recall
Semantic ranking: relevance scoring that far exceeds keyword matching

Measured Relevance Gain

Microsoft measures an average 54% improvement in answer relevance compared to traditional RAG approaches when using Foundry IQ Knowledge Bases. For methodological details, see the article Foundry IQ: Improve recall by up to 54% with knowledge bases on the Microsoft Community Hub.

Benefits by user profile

Profile	Main Benefit	What Changes Concretely
Developer / Architect	Secure connection to enterprise data	Direct selection of an existing Knowledge Base from Copilot Studio, permissions and governance automatically inherited
End User	Transparent access to organizational data	Accurate answers with inline citations, automatic respect of access rights, no configuration required
IT Administrator / Security	Centralized compliance and control	CMK, ACL, network isolation and Entra ID applied natively without operational overhead

Configuring Foundry IQ in Copilot Studio: Step-by-step guide

Prerequisites before you start

Before starting configuration, ensure you have the following:

An active Microsoft AI Foundry connection in your tenant
At least one Foundry IQ Knowledge Base already created and indexed
An existing agent created in the new Copilot Studio experience (documentation: Create an agent)
Sufficient permissions to modify agent configuration in Copilot Studio

Note — Preview

This feature is currently in public preview. Some behaviors may change before general availability. See the Copilot Studio preview documentation for the latest updates.

Configuration steps

Open Microsoft IQ configuration in your agent

Go to Copilot Studio and open the agent to configure. Navigate to the Build tab. In the components panel, select Microsoft IQ to open the Add Microsoft IQ dialog.

Select Foundry IQ as your source

In the dialog, among the available options, select Foundry IQ as the type of knowledge source to connect to your agent.

Choose your Foundry IQ connection

Select the Foundry IQ connection corresponding to your AI Foundry environment. If no connection is available, you must first establish the connection from the Microsoft AI Foundry portal.

Select the target Knowledge Base

Among the Knowledge Bases available in your Foundry IQ connection, select the one you want to use as the agent's primary knowledge source. Make sure the Knowledge Base is properly indexed and that ACLs are configured consistently with target users.

Configure agent instructions

Write or adapt your agent's system prompt instructions so it knows how to leverage the connected Knowledge Base. Example instruction:

TEXT

1You are an enterprise assistant. Answer only based on documents from the connected Knowledge Base. Always cite your sources with document references. If you cannot find the information, state this clearly.

These instructions help shape agent behavior and optimize the quality of responses generated from Foundry IQ.

Test the agent with your enterprise data

Once configuration is saved, launch a test session from the Chat panel in Copilot Studio. Ask questions related to content indexed in your Knowledge Base. Verify:

The presence of inline citations in responses
Respect of access rights (test with different user accounts)
The semantic relevance of responses relative to queries

Reference architecture: Foundry IQ in an enterprise context

For architects and developers who want to understand the complete data flow, here is a schematic representation of the integration:

📄YAML

1# Foundry IQ + Copilot Studio Architecture
2data_flow:
3  sources:
4    - SharePoint Online
5    - Azure Blob Storage
6    - SQL / Cosmos DB
7    - Custom REST APIs
8  indexing:
9    service: Foundry IQ (Azure AI Search)
10    features:
11      - Automatic semantic chunking
12      - Vectorization (Azure OpenAI embeddings)
13      - ACLs synchronized from Entra ID
14      - CMK encryption
15  agent:
16    platform: Microsoft Copilot Studio
17    component: Microsoft IQ (Foundry IQ connector)
18    retrieval_capabilities:
19      - Agentic retrieval
20      - Semantic ranking
21      - Iterative query planning
22  user:
23    interface: Microsoft 365 (Teams, SharePoint, Web)
24    permissions: Inherited from Entra ID via ACL
25    responses: Grounded with inline citations

Points of attention for IT teams

Data Governance — Critical Point

Before connecting a Knowledge Base to a Copilot Studio agent exposed to users, absolutely verify ACL configuration at each data source level. Agent responses respect Foundry IQ permissions, but only if these have been correctly mapped from the original sources. A source → Foundry IQ → Entra ID permission chain audit is strongly recommended before any production deployment.

Index updates: define a regular indexing policy (scheduled indexers) to ensure data freshness available to the agent
Monitoring: enable Azure AI Search diagnostic logs to trace queries and detect retrieval anomalies
Costs: assess inference costs related to embeddings and semantic queries based on expected usage volume
Version management: version your Knowledge Bases to allow rollbacks if response quality degrades

Resources and official documentation

To dive deeper into each aspect of this integration, see the following resources:

Connect Foundry IQ to a Copilot Studio agent (preview) — Microsoft Learn
Azure AI Search Security Overview — Microsoft Learn
Complete Foundry IQ Documentation — Microsoft AI Foundry
Create and Manage Knowledge Bases in AI Foundry — Microsoft Learn
Copilot Studio Overview — Microsoft Learn