Work IQ, Fabric IQ, Foundry IQ, Web IQ: Microsoft’s Four-Layer Brain for AI Agents, Explained

If you've felt buried under Microsoft's AI naming over the past few months, Work IQ, Fabric IQ, Foundry IQ, Agent 365, Copilot Studio, and Microsoft Foundry, you're not alone. At Build 2026, Microsoft did something genuinely useful: it organised four of those pieces into a single, coherent architecture called Microsoft IQ. Once you see the structure, the noise quiets down considerably. 

This post breaks down what Microsoft IQ actually is, what each of its four layers does, and most importantly, why this matters for any organisation building or buying AI agents on the Microsoft stack. 

The Problem Microsoft IQ Solves 

The problem has a specific name: agents keep creating data silos. An agent built on Dynamics 365 doesn't know what's happening in Teams. A custom agent built in Foundry doesn't know who the important stakeholders in the organisation are. A reporting agent built in Power BI doesn't understand the informal context behind why a number changed. 

Every AI agent needs context to be useful, and until now, every team building an agent had to wire up that context themselves: chunking documents, building retrieval pipelines, connecting to a dozen different data sources, and re-solving the same integration problem project after project. 

Microsoft IQ is Microsoft's answer: a shared, reusable intelligence substrate that any agent built in Copilot Studio, Microsoft Foundry, GitHub Copilot, or M365 Copilot can plug into, instead of building context from scratch every time. 

The Four IQ Layers, Explained 

Here's what each layer actually does in plain language, not marketing language. 

Work IQ – Organisational intelligence 

Captures semantic understanding of emails, meetings, documents, chats, and collaboration patterns inside the Microsoft 365 trust boundary. It's how an agent knows who the important stakeholders are, how a project is actually progressing, and what "normal" looks like for your organisation. This is the layer behind Microsoft Scout, the new always-on autopilot agent. 

Fabric IQ – Business data & semantics 

The semantic layer over your structured data in OneLake and Power BI. It teaches agents what "customer," "order," or "region" means in your company's specific language, not a generic interpretation. Built on semantic models, ontologies, and Graph in Fabric, so every agent in the organisation reasons about data the same way. 

Foundry IQ – Unified knowledge retrieval 

The managed knowledge layer that unifies Work IQ, Fabric IQ, Azure SQL, File Search, and MCP sources behind one SLA-backed retrieval endpoint. Instead of building a custom retrieval-augmented generation pipeline for every data source, developers point an agent at one Foundry IQ knowledge base and let it handle chunking, indexing, and permission-aware retrieval. 

Web IQ – Live external grounding 

The newest layer of AI-native grounding APIs, powered by Bing, that discover, rank, extract, and package fresh information from web pages, news, images, and video. This is what lets an agent reason about current pricing, competitive moves, or breaking news rather than relying only on internal data or a frozen training cutoff. 

How the Four Layers Connect 

The cleanest way to understand the relationship is a layered stack, not four separate products competing for attention: 

1. Foundation: Fabric IQ: Grounds agents in your structured business data, what's actually happening in the business, expressed in your company's own terms. 
2. Context: Work IQ: Adds the human layer of how people are actually working, who's involved, and what the informal organisational signals say. 
3. Reach: Web IQ: Extends context beyond your tenant's live market, news, and competitive signals an internal-only agent would never see. 
4. Retrieval: Foundry IQ: The unifying access layer. Agents query one knowledge base; Foundry IQ decides which of the other three layers (plus SQL, files, and MCP) to pull from. 

In practice, most builders won't integrate with all four layers directly. Foundry IQ is the intended integration point. You configure it with the knowledge sources relevant to your agent (Work IQ for organisational context, Fabric IQ for business data, Web IQ for live external signals), and Foundry IQ's agentic retrieval engine handles the rest: planning queries, selecting sources, running parallel searches, enforcing permissions, and returning grounded, cited answers. 

"Work IQ from Microsoft 365 provides signals on how your organization operates, Fabric IQ brings business meaning to the data in Power BI, and Foundry IQ unifies and centralizes access to knowledge to ground every agent with the right context." 

— Pablo Castro, Corporate Vice President & Distinguished Engineer, Microsoft 

A Worked Example: Supply Chain Agent 

The architecture is easier to grasp with a concrete scenario. Suppose your organisation wants an agent that helps manage supply chain delays: 

1. Fabric IQ detects the anomaly. It identifies that certain suppliers are trending late beyond historical norms, using semantic models and real-time data in OneLake expressed in terms the business already uses, like "supplier" and "lead time." 
2. Work IQ adds organisational context. It surfaces who's been discussing this supplier in Teams, which project this affects, and who the right stakeholders are to loop in based on actual collaboration patterns, not an org chart. 
3. Web IQ checks external context. It checks for live news about the supplier, a factory closure, a shipping disruption, or a regulatory change that wouldn't appear anywhere in your internal data. 
4. Foundry IQ ties it together. The agent queries one Foundry IQ knowledge base, which retrieves and synthesises across all three sources, then the agent reasons over the combined context to recommend or even take the next action. 

Why This Matters 

Without Microsoft IQ, building this agent means custom-engineering three separate integrations and a fragile glue layer between them. With it, the integration work is largely done; your team focuses on the business logic, not the plumbing. 

Rollout Timeline & Maturity 

Not all four layers are at the same stage of maturity. This matters for planning, some capabilities are safe to build on today, others are still settling. 

Practical Guidance 

If you're scoping an agent project today, Foundry IQ and Work IQ are mature enough for production planning. Fabric IQ's semantic model foundation is solid; treat Ontologies as a near-term enhancement rather than a day-one dependency. Web IQ is worth understanding architecturally now, but plan a waitlist conversation with Microsoft before committing it to a roadmap. 

Where You'll Actually Encounter This 

Microsoft IQ isn't a separate product you buy it's infrastructure that shows up inside tools you may already be using or evaluating: 

• Microsoft 365 Copilot & Cowork: increasingly grounded in Work IQ and Fabric IQ for organisational and business context 

• Copilot Studio: agents built here can tap Microsoft IQ as knowledge sources without custom integration 

• Microsoft Foundry: Toolboxes connect directly to Work IQ, Fabric IQ, and Foundry IQ, so pro-code agents get the same grounding with less plumbing 

• Agent 365: the governance and identity layer that wraps around agents using Microsoft IQ, ensuring access stays permission-aware and auditable 

• GitHub Copilot: can query Microsoft IQ context, including governed Fabric data, directly from the CLI 

In other words, if your organization is already investing in Copilot, Copilot Studio, or Foundry-based agents, you're going to interact with Microsoft IQ whether or not the name shows up explicitly in the product UI. 

What This Means for Your Organization 

For business and IT leaders evaluating an agent strategy, Microsoft IQ changes a few practical decisions: 

• Start with one use case: Pick a single high-impact scenario. Ground it in Fabric IQ, reason through Foundry IQ, and understand human impact through Work IQ. Don't try to adopt all four layers at once. 

• Governance comes first: Microsoft IQ makes more of your organisational and business data queryable by agents than ever before. Pair any rollout with Agent 365 governance and a clear data access review. 

• Budget for activation: Work IQ billing activation needs to be turned on by your M365 tenant admin. Foundry IQ pricing activation is expected later in 2026, with 30-day notice plan procurement conversations now. 

• Map your knowledge needs: Before building, map which of the four layers your agent actually needs: organizational signals (Work IQ), business data (Fabric IQ), live external context (Web IQ), or all three via Foundry IQ. 

Where Peafowl IT Fits 

This is exactly the kind of architecture decision we help clients work through. The Microsoft IQ stack is powerful, but it's also genuinely new, and the gap between "technically available" and "safely and effectively deployed" is where most agent projects stall. 

A typical Peafowl IT engagement starts by mapping your highest-value agent use case against the four IQ layers, assessing what's already GA versus what to plan around, and building a governance-first rollout plan that pairs Microsoft IQ adoption with Agent 365 controls from day one. 

Not sure where Microsoft IQ fits in your agent roadmap? 

Our Agent Architecture Readiness session maps your use case against Work IQ, Fabric IQ, Foundry IQ, and Web IQ and tells you exactly what to build first. 

Explore Our AI & Agent Services 

FAQ's