A strategic overview of Work IQ, Fabric IQ, Foundry IQ, and Web IQ — and why the architecture changes everything for enterprise AI.

For the past three years, enterprises have been deploying AI tools one layer at a time. A Copilot here, an analytics model there, a custom agent built by a developer team somewhere else. Each deployment was its own self-contained project. Each agent started from scratch, with no shared understanding of the organisation's data, its people, or how work actually gets done.

That is the isolation problem. And it is why so many AI deployments have delivered less than expected. Not because the models are weak. Because the models have no context.

Microsoft's answer to that problem is the IQ stack: a unified intelligence architecture announced at Ignite 2025 and brought to general availability at Build 2026. It is one of the most significant architectural shifts in enterprise software in years, and understanding it is now essential for any organisation running a serious Microsoft AI programme.

The differentiator for enterprise agents is not which model they call. It is what they know about your organisation.

The Problem the IQ Stack Is Solving

Enterprise AI has a context gap. It is the structural inability of AI tools to maintain a persistent, accurate understanding of the organisation they are operating inside.

A support agent built on Azure AI Foundry does not know who the key account contacts are. A Copilot embedded in Microsoft 365 does not know the business definitions your finance team uses for revenue. An agent built in Copilot Studio does not know which project has the highest strategic priority this quarter. Each tool has access to some data. None of them has a coherent picture.

The result is agents that hallucinate business context, give answers that are technically correct but practically wrong, and require constant manual correction. This is not a model quality problem. It is an architecture problem.

The IQ stack addresses it directly. Rather than leaving each agent to build its own context from scratch, Microsoft has defined a shared intelligence layer that all agents can draw from. Build the context once, make it available everywhere.

The Four Components of Microsoft IQ

Microsoft IQ comprises four components, each addressing a different domain of enterprise context. They are designed to work together, though each delivers value independently.

Work IQ: Understanding How Your Organisation Operates

Work IQ is the intelligence layer built on top of Microsoft 365. It captures the signals generated by everyday collaboration: emails, Teams conversations, meeting transcripts, documents, calendar patterns, and organisational relationships.

The practical capability is significant. An agent equipped with Work IQ does not just know what is in a SharePoint document. It knows who wrote it, who the relevant stakeholders are, how recent it is, what related discussions have taken place, and what the current status of the project it belongs to is. It understands organisational context, not just content.

Work IQ is built on Microsoft Graph and extends it with memory capabilities that learn and retain patterns over time. This is what makes it possible for agents to say "the team lead on this account is usually copied on decisions like this" rather than simply retrieving a file.

The Work IQ APIs reached general availability on 16 June 2026, making this context programmable and accessible to developers building on Microsoft Foundry and Copilot Studio.

Fabric IQ: Understanding What Your Business Data Means

Fabric IQ addresses a different problem: the semantic gap between raw data and business meaning. Most enterprise data platforms store data accurately. They do not store what the data means in the specific context of that business.

When three different business units define 'revenue' differently, every agent querying that data will produce different answers. When a 'customer' record in one system does not map cleanly to a 'client' record in another, agents operating across both will produce inconsistencies. Fabric IQ solves this through ontologies: structured definitions of business concepts, relationships, and rules that sit on top of the data in OneLake.

Define 'Customer', 'Order', or 'Shipment' once inside Fabric IQ, and every agent, report, and AI model in the environment speaks the same language. The semantic layer does not replace Power BI or existing data models. It extends them, making their business logic available to agents rather than just dashboards.

Fabric IQ is also the component that powers the operational loop: agents that can observe live business signals, reason over shared context, and take governed action in real time rather than simply reporting on historical data.

Foundry IQ: Giving Agents Access to Enterprise Knowledge

Where Work IQ understands people and process, and Fabric IQ understands structured data, Foundry IQ handles the vast body of unstructured enterprise knowledge: policies, procedures, contracts, technical documentation, institutional knowledge stored in documents and databases that do not fit neatly into a data model.

Foundry IQ is a managed knowledge layer built on Azure AI Search. It connects structured and unstructured data across Azure, SharePoint, OneLake, and the web, and makes that knowledge accessible to agents through a single, permission-aware API. Instead of every development team building their own retrieval pipeline from scratch, they can access a shared, governed knowledge base.

The traceability Foundry IQ provides is particularly valuable in regulated environments. Rather than an agent generating an answer with no audit trail, it can show exactly which documents informed the response, respect access boundaries, and operate within defined governance constraints. For industries where compliance and explainability matter, this is not a nice-to-have. It is a prerequisite.

Web IQ: Real-Time Context from Outside the Organisation

Announced at Build 2026, Web IQ is the fourth component of the stack. It extends the intelligence layer beyond the enterprise boundary by grounding agents in real-time web knowledge, powered by Bing.

The practical use case is agents that need to reason over both internal context and external signals simultaneously. A procurement agent evaluating supplier risk needs internal data on existing contracts and spend. It also needs current market intelligence on supplier stability, geopolitical factors, and pricing trends. Web IQ makes that combination possible without custom integration work for each use case.

How the Stack Fits Together

The IQ components are designed to be composable. Most enterprise agents will not need all four simultaneously. But the architecture makes it straightforward to combine them when the use case demands it.

Fabric IQ sits at the foundation, providing structured business data and semantic grounding.

Foundry IQ sits in the middle, providing knowledge retrieval across unstructured content and the ability to reason over it securely.

Work IQ sits at the top of the enterprise layer, providing organisational intelligence about people, relationships, and how work actually happens.

Web IQ extends the stack outward, adding real-time global context when the agent needs to reason beyond the enterprise boundary.

Consider a sales agent asked to prepare for a key account meeting. Work IQ provides context on the relationship history, key contacts, and recent internal communications. Fabric IQ provides structured data on revenue, contract value, and account health. Foundry IQ retrieves the relevant proposal documents, onboarding materials, and service agreements. Web IQ adds current intelligence on the client's recent announcements, competitive landscape, and market conditions. None of those context sources required a custom integration. The agent accessed them all through the shared IQ layer.

An agent built on the IQ stack does not just know what is in your data. It knows what that data means, why it matters, and who in the organisation it is relevant to.

What This Changes for Enterprise IT Leaders

The IQ stack is not primarily a developer story. Its implications run across the organisation.

1. Data strategy becomes AI strategy. The investments organisations have made in Microsoft Fabric, OneLake, and Power BI semantic models now have a direct line to AI value. The quality of the business context available to agents is determined by the quality of the data estate underneath them.

2. Governance extends to knowledge, not just data. Foundry IQ's permission-aware architecture means access controls on documents and data propagate to agents automatically. An agent cannot surface information that the user it is operating on behalf of does not have permission to see. This resolves one of the most persistent concerns about deploying AI in regulated environments.

3. The ROI case for Copilot becomes clearer. Much of the underperformance in early Copilot deployments traces back to the context gap. Agents with access to Work IQ and Fabric IQ are materially more useful because they start from an accurate understanding of the organisation rather than from scratch.

4. Development costs for custom agents fall. Foundry IQ's reusable knowledge bases mean development teams are not rebuilding retrieval infrastructure for every new agent project. The shared context layer significantly reduces the time and cost of getting a new agent from prototype to production.

The Bigger Architectural Bet

Microsoft is not simply launching new features. It is articulating an enterprise AI operating model built around a shared intelligence layer that sits beneath all AI experiences, whether those experiences are built in Copilot Studio, Microsoft Foundry, GitHub Copilot, or third-party tools.

This is a significant strategic commitment. The IQ stack is Microsoft's answer to the fragmentation that has characterised the first wave of enterprise AI deployment. Rather than every team building isolated AI experiences on top of isolated data, the architecture envisions a shared contextual foundation that makes every agent smarter without every developer having to solve the same grounding problems independently.

It also represents a clear signal about where Microsoft sees competitive advantage. The model quality race is becoming less differentiated. The context and governance layer is becoming the moat. Enterprises that build their AI programmes on a well-governed, semantically coherent data foundation will have a structural advantage over those still running isolated experiments.

The organisations that get ahead of this shift will not be the ones with the most agents. They will be the ones with the best-grounded agents, operating on the most coherent enterprise data foundation, within the most mature governance framework. The IQ stack is how Microsoft is making that possible within its ecosystem.

What Enterprises Should Be Doing Now

5. Audit your Microsoft Fabric and OneLake readiness. The IQ stack's value is directly proportional to the quality of the data and semantic models underneath it. If your data estate is fragmented, start there.

6. Map your unstructured knowledge assets. Identify the policies, procedures, and documents that agents will need to access. Foundry IQ can make these available in a governed, permission-aware way, but the content needs to be in a discoverable location.

7. Evaluate Work IQ for your highest-value agent use cases. The Work IQ APIs are now generally available. Agents that need to operate within the context of real organisational workflows should be assessed for Work IQ integration.

8. Connect your AI strategy to your data governance programme. The IQ stack makes them inseparable. The quality of enterprise AI is now a direct function of the quality of enterprise data governance.

9. Plan for a shared context layer, not siloed agent projects. The organisations that will derive the most value from the agent era are those designing for reuse and shared grounding from the outset, not retrofitting it afterwards.

Want to understand how the Microsoft IQ stack applies to your organisation's current data and AI programme?

VE3 works with enterprise teams across the UK and beyond to design and implement Microsoft AI architectures grounded in Fabric, Foundry, and the full IQ stack. Get in touch to start the conversation.