Microsoft IQ: Building Context for AI Agents Across the Enterprise

Microsoft IQ is Microsoft’s framework for giving AI agents four types of enterprise context: workplace activity, governed business data, institutional knowledge, and current external information.

Those four types of context are delivered through:

• Work IQ, which helps agents understand people, meetings, conversations, documents, responsibilities, and how work gets done.

• Fabric IQ, which gives agents a governed understanding of business data, metrics, entities, relationships, rules, and operational conditions.

• Foundry IQ, which connects agents to authoritative policies, documentation, procedures, and reusable enterprise knowledge bases.

• Web IQ, which brings current external information from the public web into the agent’s reasoning process.

Together, these capabilities address a specific limitation in enterprise AI: agents can access information without understanding how the information fits together.

A customer record may exist in a CRM system. Revenue and product usage may be defined in Microsoft Fabric. Recent customer decisions may be buried in email and Teams. Contract requirements may live in an approved knowledge base. A new regulatory announcement may only be available on the public web.

Without a shared context architecture, every agent must independently retrieve those sources, interpret their meaning, reconcile conflicting information, apply permissions, and determine which evidence should influence the answer.

Microsoft IQ is designed to make that context reusable across Copilot experiences, custom applications, and AI agents. It does not combine every source into one universal database. Instead, it provides complementary intelligence layers that agents can call based on the question, decision, or workflow in front of them.

This blog explains what each Microsoft IQ capability provides, how the four layers work together, where Agent 365 fits, and what organizations need to govern before agents can use enterprise context to take meaningful action.

Quick Answer

Microsoft IQ is Microsoft’s enterprise context layer for AI agents.

It brings together workplace activity, governed business data, institutional knowledge, and current external information so agents can understand more than the contents of a document or database record. Agents can use Microsoft IQ to interpret how information relates to people, business entities, policies, processes, operational conditions, and the world outside the organization.

Microsoft IQ is not one database, model, or application. It is a family of connected capabilities that can be used independently or together. Work IQ, Fabric IQ, Foundry IQ, and Web IQ each provide a different type of context, while Microsoft Foundry, Copilot Studio, Microsoft 365 Copilot, and Agent 365 help organizations build, deliver, govern, and operate the agents that use it.

In This Blog

• Enterprise AI Has a Context Problem

• What Is Microsoft IQ?

• The Four Context Layers of Microsoft IQ

• Work IQ: Understanding How Work Gets Done

• Fabric IQ: Understanding the State of the Business

• Foundry IQ: Building Governed Institutional Knowledge

• Web IQ: Connecting Agents to the Changing World

• How the Four IQ Layers Work Together

• What Happens When an Agent Uses Microsoft IQ?

• Why Traditional Architectures Create Context Silos

• Microsoft IQ Versus Microsoft Graph, RAG, and Semantic Models

• Where Agent 365 Fits

• Security and Governance Across Microsoft IQ

• A Practical Microsoft IQ Example

• What Microsoft IQ Does Not Solve Automatically

• How Organizations Should Prepare for Microsoft IQ

• What Microsoft IQ Means for Enterprise AI

• Frequently Asked Questions

Enterprise AI Has a Context Problem

Organizations have invested heavily in data platforms, Microsoft 365 Copilot, retrieval-augmented generation applications, analytics, knowledge management, and automation.

Those investments have created meaningful value. Employees can find information faster, summarize large volumes of content, draft communications, analyze business results, and automate tasks that previously required significant manual effort.

However, access to information is not the same as understanding it.

An agent may have access to customer records, operational data, support tickets, project documentation, policies, meeting transcripts, and business reports. It may still struggle to determine:

• Which source is authoritative

• Which information is current

• How one record relates to another

• Whether a policy applies to the current situation

• Who owns the next decision

• Which metric the company officially uses

• What happened outside the company after the model was trained

• What action is permitted

• What consequences an action may create

Humans bridge these gaps using experience and context.

An experienced employee may recognize that a delayed shipment affects a strategic customer, that the delay violates a contractual commitment, that the account team already discussed an exception, and that a new regulatory requirement changes the appropriate response.

The underlying facts may exist in four or five systems. The employee supplies the connections.

AI agents do not automatically possess that organizational understanding. If context remains fragmented, an agent may retrieve accurate facts and still reach the wrong conclusion.

Microsoft IQ is Microsoft’s attempt to make those connections explicit, governed, and reusable.

What Is Microsoft IQ?

Microsoft defines Microsoft IQ as a unified intelligence layer for enterprise AI. Its purpose is to ground agent and Copilot interactions in a shared understanding of the organization.

This represents an architectural shift in how enterprise AI systems are designed.

Historically, enterprise systems have been optimized to store, process, search, or analyze specific types of information:

• Microsoft 365 captures communication and collaboration.

• CRM systems track customers and opportunities.

• ERP platforms manage financial and operational transactions.

• Data platforms consolidate analytical information.

• Power BI semantic models define trusted business metrics.

• SharePoint and other repositories hold policies and documents.

• Search engines retrieve information from the public web.

Each system may perform its individual role well. However, the broader meaning of the business remains divided among them.

Microsoft IQ does not attempt to move every piece of information into one universal repository. Instead, it provides connected forms of intelligence that agents can use based on the task.

Microsoft announced at Build 2026 that Microsoft IQ was generally available across experiences including GitHub Copilot, Microsoft Foundry, and Copilot Studio. However, individual features within each IQ capability may have different availability, licensing, regional support, or preview status.

Organizations should review the current Microsoft IQ documentation and the product-specific documentation for the capabilities they intend to deploy.

Microsoft IQ should therefore be understood as both a product strategy and an evolving family of technical capabilities.

The Four Context Layers of Microsoft IQ

The four IQ capabilities solve different context problems.

They are not competing versions of the same product. Each layer answers a different set of questions an enterprise agent may face.

A reliable agent may use one layer or all four.

For example, an employee asking, “What did my manager say about the Q4 deadline?” primarily needs Work IQ.

An operations leader asking, “Which production orders are at risk?” may primarily need Fabric IQ.

An employee asking, “What is our international travel policy?” should receive an answer grounded through Foundry IQ.

An executive asking, “What has changed in this market over the past week?” needs Web IQ.

A strategic customer briefing may need all four.

Work IQ: Understanding How Work Gets Done

Work IQ focuses on one of the richest and most fragmented forms of enterprise context: the work itself.

Every day, organizations generate signals through:

• Outlook email

• Calendars

• Meetings

• Microsoft Teams messages

• OneDrive and SharePoint documents

• People and organizational profiles

• Collaboration patterns

• Tasks and workflows

• Connected line-of-business applications

A traditional API can retrieve an email, meeting, or file. However, the agent may still need to determine which information matters to the task and how the resources relate.

Microsoft describes Work IQ as the intelligence layer behind how work gets done. It processes workplace signals to create a semantic understanding of people, roles, organizational structure, collaboration patterns, content, and business activity.

From Workplace Data to Workplace Context

Suppose a user asks:

What changed on the Atlas project this week, and what do I need to do next?

The relevant context may be distributed across:

• A revised project plan in SharePoint

• A meeting transcript

• A Teams discussion

• An email from the customer

• A calendar event

• Several assigned action items

• The organizational roles of the people involved

A conventional retrieval system may locate documents containing “Atlas.”

Work IQ is intended to help the agent assemble a more useful picture:

• Which Atlas project the user means

• Which files were recently active

• Which meeting contained the latest decision

• Which email altered the timeline

• Which people are responsible

• What work remains incomplete

• Which actions are available to the agent

The Work IQ API Architecture

Microsoft organizes the Work IQ APIs into four domains:

1. Chat: Programmatic access to Microsoft 365 Copilot responses and agents

2. Context: Agent-ready workplace context and source information

3. Tools: Actions across Microsoft 365 resources

4. Workspaces: Persistent storage for longer-running agent tasks

The Tools layer uses a compact set of generic operations rather than requiring an agent to load hundreds of specialized Microsoft 365 tools into its context. Microsoft’s Work IQ API announcement explains how these capabilities give developers access to Microsoft 365 context and actions through an agent-oriented interface.

Workspaces provide tenant-boundary storage for intermediate files, progress, memory, and outputs during multistep work. This is valuable for agents performing tasks that cannot be completed in one interaction.

The Intelligence Behind Work IQ

Work IQ can be understood through three intelligence layers:

1. Data: Microsoft 365 resources and signals, including mail, meetings, files, people, and collaboration activity

2. Memory: Information about the user’s preferences, relationships, recurring patterns, and prior activity

3. Inference: The reasoning used to assemble the most relevant workplace context for the current request

This allows the system to move beyond literal keyword matching.

An agent may recognize that a recently edited document, a meeting attended by the user’s manager, and a Teams conversation involving the project owner are more relevant than older files that happen to contain the same project name.

How Microsoft Graph Supports Work IQ

Microsoft Graph remains the API foundation for accessing Microsoft 365 resources such as users, email, calendars, files, and Teams.

However, Microsoft Graph and Work IQ serve different roles.

Microsoft Graph may retrieve the meeting, email, or file. Work IQ helps determine why it matters to the person and task.

Why Work IQ Matters

Many business decisions never appear in a formal system of record.

They are made in meetings, clarified in chats, revised in email, and reflected in the documents people choose to work on.

Work IQ gives agents access to that dynamic workplace layer. It allows an agent to reason not only over what the organization stores, but also over how people are actively moving work forward.

Work IQ requests operate in the context of the signed-in user. Microsoft states that Work IQ respects Microsoft 365 permissions and sensitivity labels and keeps work data within the Microsoft 365 trust boundary.

Fabric IQ: Understanding the State of the Business

Work IQ explains how work is progressing. Fabric IQ explains what is happening across the business.

It is grounded in analytical, operational, and real-time data, but its purpose goes beyond giving an agent access to tables.

Organizations do not naturally think in terms of database schemas. They think in terms of:

• Customers

• Products

• Assets

• Suppliers

• Facilities

• Orders

• Shipments

• Contracts

• Risks

• Business objectives

Fabric IQ elevates technical data into this business language so people and agents can reason in terms of shared concepts rather than isolated records.

Microsoft defines three connected layers within Fabric IQ:

1. Unified data through OneLake

2. Business intelligence through Power BI semantic models

3. Operational intelligence through ontology

OneLake Provides the Data Foundation

OneLake creates a unified, governed data foundation across Microsoft Fabric.

Through ingestion, mirroring, and shortcuts, organizations can make data from multiple platforms available without requiring every workload or agent to maintain its own separate copy.

However, consolidated data is not automatically understandable. The next layers define what that information means.

Semantic Models Define Trusted Analytical Meaning

Power BI semantic models define:

• Measures

• Calculations

• Dimensions

• Hierarchies

• Relationships

• Time intelligence

• Security rules

• Approved KPI definitions

A semantic model can tell an agent how the company calculates revenue, margin, active customers, inventory turnover, or on-time delivery.

This is essential because different departments may otherwise calculate the same metric differently.

Ontology Defines Enterprise Meaning

Fabric IQ ontology extends beyond analytical calculations.

An ontology can define:

• Entity types such as Customer, Order, Product, Supplier, or Asset

• Properties associated with those entities

• Relationships among entities

• Rules and constraints

• Connections to OneLake data

• Actions available in a business scenario

Microsoft describes Fabric IQ Ontology as a shared, machine-understandable vocabulary of the business. It allows people and agents to reason consistently across domains and supports business-level questions that would otherwise require knowledge of individual tables and joins.

For example, an organization may define a Customer entity that connects:

• CRM account information

• Orders

• Invoices

• Support cases

• Product usage

• Contracts

• Customer health metrics

The agent can then reason about the customer as a business concept instead of separately interpreting every underlying system.

Graph Makes Relationships Queryable

Ontology declares which concepts connect and why. Graph stores and traverses instances of those relationships.

This can help an agent answer questions such as:

• Which customer orders depend on the affected supplier?

• Which assets share a failing component?

• How could a delay at one facility affect downstream shipments?

• Which contracts are connected to an operational risk?

Relationships become explicit, governed objects rather than logic hidden inside individual applications and queries.

Agents Operationalize the Context

Fabric data agents can answer natural-language questions over governed Fabric sources.

Operations agents can monitor changing conditions, reason over business concepts, and recommend or trigger approved responses.

Together, these capabilities move Fabric IQ beyond reporting. The same semantic foundation can support analysis, operational monitoring, agent reasoning, and coordinated action.

Why Fabric IQ Matters

Leaders do not make decisions based on isolated rows and columns. They make decisions based on customers, products, risks, assets, facilities, orders, and outcomes.

Fabric IQ creates a governed representation of those business concepts that can be reused across analytics, applications, employees, and AI agents.

Foundry IQ: Building Governed Institutional Knowledge

Not every important enterprise question is answered by workplace activity or structured business data.

Organizations also depend on institutional knowledge, including:

• Policies

• Manuals

• Procedures

• Product documentation

• Legal guidance

• Technical standards

• Support articles

• Research

• Contracts

• Training materials

• Industry playbooks

This information is often distributed across SharePoint, Azure storage, OneLake, search indexes, websites, and specialized repositories.

Foundry IQ provides a managed knowledge layer for connecting these sources to enterprise agents.

Knowledge Sources

A knowledge source connects Foundry IQ to an indexed or remote repository.

Supported scenarios include sources such as:

• Azure Blob Storage

• SharePoint

• OneLake

• Existing Azure AI Search indexes

• Public web content

Depending on the source and configuration, content may be indexed into Azure AI Search or retrieved remotely when the agent submits a question.

Knowledge Bases

A Foundry IQ knowledge base brings one or more sources together into a reusable knowledge domain.

For example, an organization could create:

• An HR policy knowledge base

• A technical support knowledge base

• A sales enablement knowledge base

• A legal and compliance knowledge base

• A manufacturing procedures knowledge base

Multiple agents can use the same knowledge base. This reduces the need for every development team to create separate ingestion, indexing, retrieval, permission, and citation systems over the same content.

Agentic Retrieval

Foundry IQ uses agentic retrieval to process complex questions.

Instead of sending one unchanged query to one search index, the retrieval engine can:

1. Interpret the request

2. Break it into subqueries

3. Select appropriate sources

4. Run keyword, vector, or hybrid searches

5. Process searches in parallel

6. Apply semantic reranking

7. Combine the evidence

8. Return supporting source references

A Foundry agent can access the knowledge base through Model Context Protocol. The knowledge base performs the retrieval work and sends the resulting evidence back to the agent. Microsoft explains this integration in its Foundry IQ connection guidance.

Indexed and Remote Retrieval

Foundry IQ can use both indexed and remote knowledge sources.

A single knowledge base may combine several sources and retrieval approaches.

Why Foundry IQ Matters

Foundry IQ creates a separation between the knowledge and the agent.

The knowledge base governs:

• Which sources are included

• How retrieval works

• Which users may access content

• How evidence is cited

• How content is maintained

The agent governs:

• How the evidence is used

• Which task is performed

• How the response is presented

• Which actions may follow

This architecture allows multiple agent experiences to work from the same governed body of institutional knowledge.

Web IQ: Connecting Agents to the Changing World

The first three IQ capabilities focus largely on context connected to the enterprise.

Web IQ connects agents to information outside the organization.

A model may understand the general history of an industry or company, but it does not automatically know:

• What happened this morning

• Which executive currently holds a role

• What a competitor recently announced

• Whether public guidance changed

• Which regulation was just published

• What current news coverage says

• Whether a product is presently available

• How an external event is developing

Web IQ is Microsoft’s AI-native grounding layer for connecting agents to fresh external evidence. It builds on Bing’s search infrastructure but is designed around the way AI systems retrieve, rank, cite, and reason over information.

Web Search Versus Agent Grounding

A conventional search engine returns links for a person to investigate.

An agent grounding system must deliver information in a form that a model can efficiently use.

That requires attention to:

• Freshness

• Relevance

• Source authority

• Passage selection

• Citations

• Latency

• Token consumption

An agent may also search several times during one task.

For example, a competitor-research agent may begin with a product announcement, discover a regulatory concern, search for the relevant rule, compare several news reports, and then locate the company’s official filing.

Web IQ makes this iterative retrieval part of the agent’s reasoning loop.

How Enterprises Can Use Microsoft Web Grounding

Microsoft web-grounding capabilities are available through experiences such as:

• Microsoft Foundry Agent Service

• Microsoft Foundry Responses API

• Microsoft 365 Copilot

• Microsoft 365 Copilot Chat

• Copilot Studio

• Bing Custom Search-based grounding

The web search tool in Microsoft Foundry Agent Service allows agents to retrieve current public information and return cited responses.

Microsoft also supports custom web search for cases where organizations want to limit an agent to approved domains or specific sections of the public web.

General Versus Restricted Web Grounding

Why Web IQ Requires Different Governance

The open web is not a governed enterprise repository.

Sources may be:

• Incorrect

• Outdated

• Biased

• Duplicated

• Manipulated

• Malicious

• In conflict with one another

Web-grounded content should be treated as untrusted input. Important claims may require corroboration, primary sources, approved domain restrictions, or human verification.

Web IQ adds current external evidence. It does not make every public source authoritative.

How the Four IQ Layers Work Together

The most valuable Microsoft IQ scenarios appear when an agent combines several kinds of context.

Consider a customer-renewal agent.

Work IQ Can Explain the Relationship

It can retrieve:

• Recent customer emails

• Upcoming meetings

• Internal Teams discussions

• Account-team members

• Previous commitments

• Active documents

• Unresolved action items

Fabric IQ Can Explain the Account

It can provide:

• Contract value

• Product usage

• Support metrics

• Renewal probability

• Revenue

• Account health

• Business relationships

Foundry IQ Can Explain the Rules

It can retrieve:

• Approved pricing policies

• Renewal procedures

• Discount authority

• Legal guidance

• Product documentation

• Customer-success playbooks

Web IQ Can Explain the External Situation

It can identify:

• Recent customer announcements

• Leadership changes

• Financial results

• Market developments

• Public security incidents

• Regulatory updates

• Competitor activity

The resulting agent can produce a much more useful briefing than an agent connected to only one system.

It can understand what employees know, what the business data shows, what the organization permits, and what has changed outside the company.

That is the larger architectural promise of Microsoft IQ.

What Happens When an Agent Uses Microsoft IQ?

Microsoft IQ is not a single pipeline that every request follows in exactly the same way.

The execution path depends on the agent, the available IQ capabilities, the user’s permissions, and the question.

A simplified runtime may look like this:

1. A User or Agent Submits a Request

The request may come through:

• Microsoft 365 Copilot

• Copilot Studio

• Microsoft Foundry

• GitHub Copilot

• A custom application

• Another agent

2. Identity and Permissions Establish the Boundary

The system determines:

• Who is making the request

• Which agent is involved

• What data the user or agent may access

• Which tools the agent may invoke

• Which policies apply

3. The Orchestration Layer Identifies the Required Context

A question about a recent meeting may use Work IQ.

A question about current sales performance may use Fabric IQ.

A policy question may use Foundry IQ.

A question about a new regulation may use Web IQ.

A complex task may use all four.

4. The IQ Layers Retrieve and Structure Context

Each capability provides agent-ready evidence in a form appropriate to its domain.

This may include:

• Workplace context

• Semantic metrics

• Ontology entities and relationships

• Document passages

• Citations

• Current web evidence

• Available actions

5. The Agent Reasons Over the Context

The agent interprets the retrieved information according to its instructions and the user’s request.

It may:

• Summarize

• Compare

• Identify risk

• Recommend an action

• Ask a follow-up question

• Invoke another tool

• Delegate work to another agent

6. Tools Support Action

Depending on its permissions, the agent may:

• Draft an email

• Create a document

• Schedule a meeting

• Update a resource

• Trigger a workflow

• Request human approval

7. Governance Systems Observe and Control the Agent

Agent 365, Microsoft Entra, Microsoft Purview, Microsoft Defender, and platform-specific administration tools can help govern identity, data access, tool use, lifecycle, monitoring, and security.

This is important because context and action create different levels of risk.

An agent that summarizes a policy is not equivalent to one that changes a customer record or sends an external communication.

Why Traditional Architectures Create Context Silos

Most enterprise platforms are optimized for specific business functions.

Individually, these platforms provide value.

Collectively, they create silos of understanding.

Humans bridge those silos using experience, memory, judgment, and relationships. They know that a record in one system relates to a conversation in another and that both are governed by a policy stored somewhere else.

Traditional AI architectures frequently reproduce the same fragmentation.

Each agent may have:

• Its own prompt

• Its own retrieval index

• Its own source connections

• Its own security logic

• Its own interpretation of business terms

• Its own action tools

• Its own version of organizational knowledge

As the number of agents grows, those independent implementations can drift.

Two agents may use different sources, apply different definitions, or provide inconsistent answers to the same question.

Microsoft IQ points toward an architecture in which important forms of context can be built, governed, and reused across many agents.

Microsoft IQ Versus Microsoft Graph, RAG, and Semantic Models

Microsoft IQ does not replace the technologies that provide data, search, semantics, and retrieval.

It builds on and connects them.

Microsoft IQ Does Not Replace Microsoft Graph

Microsoft Graph continues to provide access to resources such as users, email, calendars, files, and Teams.

Work IQ builds on workplace signals and intelligence to provide agent-ready context rather than requiring every application to assemble raw Graph data independently.

Fabric IQ Does Not Replace Semantic Models

Semantic models remain essential for governed analytical definitions.

Fabric IQ expands that context through ontology, graph relationships, operational intelligence, and agents.

Foundry IQ Does Not Replace RAG

Foundry IQ is an enterprise-oriented, managed knowledge architecture that uses retrieval patterns, including agentic retrieval.

Organizations may still build custom RAG systems when they need complete control over chunking, embeddings, infrastructure, ranking, or non-Microsoft platforms.

Web IQ Does Not Replace Direct APIs

A dedicated API is often the best source for exact, transactional, or domain-specific information.

Weather, market prices, shipment tracking, reservation availability, and account balances should generally come from an authoritative API when one is available.

Where Agent 365 Fits

Agent 365 is not a fifth Microsoft IQ context layer.

It is Microsoft’s control plane for observing, governing, and securing agents across the organization.

Microsoft describes Agent 365 through three major functions:

• Observe: Maintain visibility into agent inventory, usage, activity, performance, and risk signals.

• Govern: Apply lifecycle management, access controls, ownership, policies, and compliance requirements.

• Secure: Extend Microsoft Entra, Microsoft Purview, and Microsoft Defender protections to agents.

This distinction is important.

Microsoft IQ provides context that helps agents understand.

Agent 365 provides controls that help the organization manage the agents using that context.

As agents move from answering questions to performing actions, governance becomes increasingly important. Organizations need to know:

• Which agents exist

• Who owns them

• Which identities they use

• What information they can access

• Which tools they can invoke

• Which actions they have taken

• Whether their behavior remains within policy

• When an agent should be updated or retired

Microsoft Entra provides identity and access controls for agents. Microsoft Purview helps address data protection and compliance. Microsoft Defender adds threat detection and runtime protection. Agent 365 brings these controls into a centralized agent-management approach.

Security and Governance Across Microsoft IQ

A shared intelligence layer does not eliminate the need for governance. It makes governance more important.

The more context an agent can use, the more carefully organizations must manage identity, access, source quality, actions, and oversight.

Identity

Every agent interaction should have a clear identity model.

An agent may:

• Act on behalf of a signed-in user

• Operate as an application

• Use its own managed agent identity

The correct model depends on the agent’s purpose and risk.

Permission Enforcement

Microsoft IQ does not grant every agent universal access to enterprise information.

Work IQ respects Microsoft 365 permissions and sensitivity labels.

Fabric IQ uses Fabric, OneLake, semantic-model, and source-level security.

Foundry IQ can use source permissions, document-level access controls, Microsoft Entra identities, Azure RBAC, SharePoint ACLs, and Purview labels, depending on the source.

Web IQ uses public external information but introduces separate privacy and data-boundary considerations.

Source Governance

Permissions answer whether a user may see information.

They do not determine whether the information is correct.

Organizations still need to manage:

• Source ownership

• Data quality

• Document freshness

• Duplicate content

• Superseded policies

• Conflicting definitions

• Missing metadata

• Unclear authority

Tool and Action Governance

Reading context and changing a system are not the same level of risk.

An agent that assists by summarizing or drafting can usually operate with lighter controls than an agent that updates systems of record, changes permissions, submits requests, or moves money.

Organizations should classify agents based on:

• The sensitivity of the data they use

• The systems they can access

• The tools they can invoke

• The reversibility of their actions

• The business impact of an error

• Whether human review is required

Human Approval

Organizations should consider human approval before agents:

• Send external communications

• Update customer or employee records

• Change financial information

• Modify deadlines or contractual commitments

• Share sensitive documents

• Trigger operational processes

• Make legal, employment, or security decisions

Human oversight should be designed into the workflow rather than added only after a problem occurs.

Privacy and External Search

Web-grounded agents require particular attention to query privacy and data boundaries.

Organizations should determine:

• What information may contribute to an external search query

• Which agents may use public web search

• Whether searches should be limited to approved domains

• Which regulatory obligations apply

• How generated queries and retrieved sources are logged

Confidential business information, secrets, sensitive personal information, and unnecessary customer data should not be included in external searches.

A Practical Microsoft IQ Example

Consider a manufacturer that wants an agent to identify customer orders at risk and coordinate an appropriate response.

The information required for the decision is distributed across:

• ERP orders

• Inventory

• Supplier systems

• Production data

• Sensor events

• CRM records

• Customer emails

• Teams discussions

• Meeting transcripts

• Operating procedures

• Contractual service requirements

• Public supplier announcements

• News and regulatory information

The operations leader asks:

Which priority customer orders are at risk, why are they at risk, and what should we do next?

Fabric IQ Establishes the Business Situation

Fabric IQ can connect:

• Customer

• Sales Order

• Product

• Material

• Supplier

• Production Line

• Shipment

• Facility

Semantic models provide trusted measures such as available inventory, production capacity, backlog, customer priority, and on-time delivery.

Ontology and graph relationships help the agent trace how a supplier problem affects materials, products, production orders, shipments, and customers.

Work IQ Adds the Human and Workflow Context

Work IQ can identify:

• Recent supplier communications

• Internal Teams discussions

• Decisions made in production meetings

• Employees responsible for the affected account

• Existing action items

• Upcoming customer meetings

• A recent commitment made by the sales team

Foundry IQ Adds Approved Guidance

Foundry IQ can retrieve:

• Supplier-disruption procedures

• Customer-escalation policies

• Contractual service-level guidance

• Approved communication templates

• Safety requirements

• Alternative-sourcing procedures

Web IQ Adds External Developments

Web IQ can identify:

• A supplier’s public disruption announcement

• News coverage of the incident

• A newly published regulatory notice

• Port, transportation, or weather disruptions

• Current market availability for replacement materials

The Agent Produces a Contextual Response

The response may explain:

• Which orders are at risk

• Which business relationships create the risk

• What internal teams already know

• Which customers have been promised a response

• Which policy applies

• Which external event contributed to the issue

• What actions are available

• Which decisions require approval

The agent may then draft a customer communication, create an internal briefing, or initiate an approved escalation workflow.

No single data source could provide this complete picture.

The value comes from combining several forms of context around one business decision.

What Microsoft IQ Does Not Solve Automatically

Microsoft IQ can provide a strong architecture for agent context. It does not automatically make the underlying environment accurate, secure, or ready for autonomous action.

It Does Not Fix Poor Data Quality

If source data is duplicated, incomplete, or inconsistent, agents may reproduce those problems.

It Does Not Determine Business Authority

The organization must still decide which metric, policy, document, or system is authoritative.

It Does Not Correct Oversharing

If users already have access to information they should not see, permission-aware AI may make that existing access easier to exercise.

It Does Not Capture Undocumented Knowledge

If important rules exist only in employees’ heads, the agent cannot reliably use them until they are documented or represented in a system.

It Does Not Eliminate Hallucinations

Grounding and citations reduce unsupported responses, but agents still require evaluation, clear instructions, source review, and appropriate abstention behavior.

It Does Not Make Public Web Content Trustworthy

Web IQ improves access to current evidence. It does not guarantee that every retrieved source is correct.

It Does Not Decide How Autonomous an Agent Should Be

The organization must determine when an agent may act, when it should ask for approval, and when it must escalate to a person.

How Organizations Should Prepare for Microsoft IQ

Organizations do not need to deploy all four IQ capabilities at once.

A more practical approach is to begin with a valuable business decision and identify the context required to improve it.

1. Start with a Decision or Workflow

Avoid beginning with a vague goal such as “connect all enterprise knowledge.”

Start with a specific outcome, such as:

• Prepare an account briefing

• Identify production orders at risk

• Answer employee policy questions

• Coordinate a security incident

• Support customer-service agents

• Evaluate a vendor

• Accelerate project handoffs

2. Map the Context Required

Determine which questions the agent must answer.

For example:

• What happened?

• Who is involved?

• Which business entities are affected?

• What does the data show?

• Which policy applies?

• What changed externally?

• What actions are available?

• Who must approve the next step?

This reveals which IQ capabilities are actually needed.

3. Review Data and Content Authority

For every required context source, identify:

• The system of record

• The business owner

• The approved definition

• The review date

• The intended audience

• The security classification

• The update frequency

• Conflicting or duplicate sources

4. Strengthen Semantic Models

Review Power BI semantic models for:

• Clear business terminology

• Trusted measures

• Consistent dimensions

• Appropriate relationships

• Row-level security

• Active business ownership

• Suitability for agent use

Semantic quality remains foundational even when ontology is added.

5. Begin Modeling Business Entities and Relationships

Identify the entities central to the use case.

Examples might include:

• Customer

• Account

• Contract

• Product

• Order

• Supplier

• Asset

• Facility

• Employee

• Project

Then document the relationships, rules, constraints, and actions that give those entities meaning.

6. Improve Microsoft 365 Governance

Review:

• SharePoint permissions

• Teams membership

• OneDrive sharing

• Guest access

• Anonymous links

• Sensitivity labels

• Content retention

• Stale sites

• Duplicate files

• Organizational profile data

Work IQ can only be as trustworthy as the Microsoft 365 environment it interprets.

7. Build Governed Knowledge Domains

Organize institutional knowledge into clearly owned domains rather than one massive enterprise index.

For each domain:

• Identify authoritative sources

• Remove obsolete content

• Improve metadata

• Define permissions

• Establish refresh schedules

• Create representative test questions

• Decide how the agent should handle missing evidence

8. Define External-Search Policies

Decide:

• Which agents may use Web IQ

• Which questions justify web search

• Whether domains should be restricted

• Which sources should receive priority

• When claims require corroboration

• What information must never influence an external search query

• When a person must review the result

9. Design Agent Identity and Permissions

Determine whether the agent should:

• Act on behalf of a user

• Use application permissions

• Operate under its own identity

• Have read-only access

• Use write tools

• Require approval for selected actions

Apply least privilege from the beginning.

10. Evaluate the Complete System

Do not measure only whether the model produced an answer.

Evaluate:

• Retrieval relevance

• Metric accuracy

• Source authority

• Freshness

• Citation quality

• Permission enforcement

• Relationship reasoning

• Action selection

• Tool reliability

• Abstention

• Human escalation

• Latency

• Cost

• Business outcomes

11. Establish Agent Lifecycle Governance

Every production agent should have:

• A named owner

• A defined purpose

• Approved data sources

• Documented tools and permissions

• A risk classification

• Evaluation requirements

• Monitoring

• Change management

• A retirement process

Agent 365 can help organizations centralize many of these controls as agent adoption grows.

What Microsoft IQ Means for Enterprise AI

The first phase of generative AI focused heavily on model capability.

The next phase will depend on whether organizations can give those models the right context.

An enterprise agent needs to understand:

• What people are working on

• What the business data means

• How entities relate

• Which rules apply

• What the organization officially knows

• What has changed in the outside world

• Which actions are permitted

• When a person must remain responsible

Microsoft IQ brings these forms of understanding into one broader enterprise intelligence strategy.

Work IQ captures the context of work.

Fabric IQ captures the context of the business.

Foundry IQ captures the context of institutional knowledge.

Web IQ captures the context of the external world.

Agent 365 surrounds the growing agent ecosystem with identity, observability, governance, security, and lifecycle management.

The larger opportunity is not to give every agent access to every system.

It is to create shared, governed context that agents can use consistently.

That context can become an enterprise asset in the same way that data platforms, semantic models, APIs, and identity systems became shared infrastructure. Instead of rebuilding the organization’s meaning for every agent, teams can define important concepts, sources, relationships, policies, and controls once and make them available across many AI experiences.

Organizations that begin this work now will be better positioned to move beyond isolated copilots and proofs of concept. They will be able to build agents that understand the organization well enough to support real decisions and workflows without sacrificing governance.

Emergent Software helps organizations prepare their Microsoft environments for enterprise AI, from Microsoft 365 and Fabric foundations to Microsoft Foundry, Copilot Studio, Azure AI Search, security, and custom agent development.

Explore Emergent Software’s Copilot and custom AI agent development services or contact our team to discuss how Microsoft IQ can support a more contextual, governed, and scalable enterprise AI strategy.

Frequently Asked Questions

What is Microsoft IQ?

Microsoft IQ is Microsoft’s unified intelligence layer for enterprise AI. It combines Work IQ, Fabric IQ, Foundry IQ, and Web IQ so agents can use workplace context, business context, institutional knowledge, and current external information.

Is Microsoft IQ a single product?

No. Microsoft IQ is a family of interconnected capabilities available across Microsoft’s agent and Copilot ecosystem. Organizations can use individual IQ capabilities independently or combine them based on the context an agent requires.

What are the four components of Microsoft IQ?

The four components are Work IQ, Fabric IQ, Foundry IQ, and Web IQ. Work IQ explains how work happens, Fabric IQ explains the state of the business, Foundry IQ provides governed institutional knowledge, and Web IQ supplies current external evidence.

What is Work IQ?

Work IQ is Microsoft’s workplace intelligence layer. It combines emails, meetings, documents, chats, people, collaboration patterns, and organizational activity so agents can understand how work gets done and take appropriate actions across Microsoft 365.

What is Fabric IQ?

Fabric IQ is Microsoft Fabric’s business-context layer. It combines OneLake, Power BI semantic models, ontology, graph, data agents, and operations agents so people and AI systems can reason over trusted business concepts, data, relationships, and operational conditions.

What is Foundry IQ?

Foundry IQ is Microsoft’s managed enterprise-knowledge layer. It allows organizations to create reusable knowledge bases from sources such as SharePoint, Azure Blob Storage, OneLake, Azure AI Search, and public web content, with agentic retrieval, permissions, and citations.

What is Web IQ?

Web IQ is Microsoft’s external grounding layer for AI agents. It uses Microsoft’s web-search infrastructure to retrieve fresh, citation-ready evidence from the public web so agents can answer questions involving current events, announcements, regulations, markets, and other changing information.

How is Microsoft IQ different from Microsoft 365 Copilot?

Microsoft 365 Copilot is an end-user AI experience across Microsoft 365. Microsoft IQ provides context that Copilot and other agents can use. Copilot is one experience layer, while Microsoft IQ is part of the intelligence and grounding architecture beneath agent experiences.

Does Microsoft IQ replace Microsoft Graph?

No. Microsoft Graph continues to expose Microsoft 365 resources and data through APIs. Work IQ builds on Microsoft 365 signals and intelligence to provide agent-ready workplace context, reasoning, and tools.

Does Fabric IQ replace Power BI semantic models?

No. Semantic models remain central to Fabric IQ because they define trusted measures, dimensions, relationships, and analytical logic. Fabric IQ extends that foundation with ontology, graph, operational intelligence, and agent grounding.

Does Foundry IQ replace RAG?

Not entirely. Foundry IQ provides a managed, reusable enterprise architecture for RAG and agentic retrieval scenarios. Organizations may still use custom RAG when they require specialized infrastructure, ranking, embedding, or retrieval control.

Where does Agent 365 fit into Microsoft IQ?

Agent 365 is not one of the four Microsoft IQ context layers. It is Microsoft’s control plane for observing, governing, and securing agents. It works with Microsoft Entra, Microsoft Purview, Microsoft Defender, and Microsoft 365 administration capabilities.

Can one agent use all four Microsoft IQ capabilities?

Yes. An agent can use several context sources when the scenario requires them. For example, a customer-planning agent could use Work IQ for communications, Fabric IQ for account metrics, Foundry IQ for approved playbooks, and Web IQ for current external developments.

Does Microsoft IQ automatically give agents access to all enterprise data?

No. Access depends on the user, agent identity, permissions, security policies, connected sources, and the configuration of each capability. Organizations remain responsible for least-privilege access, source governance, and agent controls.

Is Microsoft IQ generally available?

Microsoft announced Microsoft IQ as generally available across experiences including GitHub Copilot, Microsoft Foundry, and Copilot Studio at Build 2026. However, individual capabilities and integrations have different release statuses, licensing requirements, and regional availability. Organizations should review current Microsoft documentation before production deployment.

Where should an organization start with Microsoft IQ?

Start with one valuable business decision or workflow. Identify the workplace, business, institutional, and external context the agent needs, then improve those specific sources, permissions, definitions, and relationships. Test the complete workflow before expanding to additional agents or domains.