A few weeks ago, I found myself staring at a slide full of new Microsoft AI names and thinking… wait a second. Work IQ. Fabric IQ. Foundry IQ. Agent 365. Agent Factory. And that’s before we even get into Copilot Studio, Copilot Studio Light, and Microsoft Foundry. If you’re a smart technical leader and you’re feeling a little overwhelmed, you’re not alone. I was confused too. The pace of new terms and announcements over the past few months has been fast. Very fast. This post is my attempt to slow it down and simplify what’s actually happening underneath the names.

Here’s the mental model that helped me. There are three capability umbrellas (also called IQ layers or architecture labels). Underneath them are build and orchestration tools. And wrapping around all of it is a runtime and governance layer that makes AI safe and scalable. Once you see that structure, the noise quiets down and the pieces start to fit together.

Let’s start with the three IQ layers, because these are not products you buy. They describe what intelligence you are trying to achieve as an organization.

You won’t find “Work IQ” or “Foundry IQ” as standalone SKUs or admin blades. I’m using them here as architectural labels to describe Microsoft’s direction, not official product names. They are the intelligence outcomes Microsoft wants organizations to achieve.

Work IQ

Work IQ is intelligence about how work actually happens inside your company. It models the way people collaborate — their documents, meetings, conversations, relationships, and day-to-day interaction patterns across Microsoft 365. It understands meetings, chats, emails, documents, collaboration signals, overload patterns, and exception handling. It’s powered by Microsoft 365 data and organizational graph signals.

When Work IQ is functioning well, you can answer questions like: Which teams are buried in manual exception handling? Where are email threads exploding? Who is spending ten hours a week chasing status updates?

But it goes a step further.

This intelligence layer becomes the personalization engine for AI agents. Instead of a generic assistant, Work IQ helps agents understand who you work with, what you work on, and how you get things done. It brings user context into the equation — not just data context.

For architects, Work IQ is effectively the user-context layer. It’s the bridge between user behavior and the actions agents take inside the enterprise environment. It connects human workflow signals to intelligent action.

It’s intelligence about human workflow — not just data.

Fabric IQ

Fabric IQ is intelligence grounded in structured enterprise data. It lives in Microsoft Fabric and understands your KPIs, trends, anomalies, lineage, and semantic models. But more importantly, Fabric IQ sits on top of the enterprise data estate and injects business meaning into it through a semantic layer and ontology. It turns raw datasets into entities, relationships, metrics, timelines, and operational context that agents can reason about.

This is the analytical brain.

It doesn’t just see tables. It sees business reality: customers, inventory, regions, exceptions, dependencies, patterns. It recognizes supply chain bottlenecks, revenue shifts, production throughput constraints, and operational performance trends because the data has been modeled with intent.

Fabric IQ is what allows AI to reason over trusted, governed, enterprise-grade data (structured, semi-structured, and unstructured) instead of guessing from disconnected spreadsheets.

From a solution architecture view, Fabric IQ becomes the data-context layer — essential for designing analytics, decision systems, and operational AI pipelines. It’s where analytics maturity meets AI capability.

And for DevOps teams, Fabric IQ introduces a new discipline: semantic models must be versioned, governed, deployed, and monitored the same way we handle code. Because once agents depend on business meaning, that meaning becomes production infrastructure.

In short, Fabric IQ gives AI structured business context — not just raw data.

Foundry IQ

Foundry IQ is intelligence grounded in knowledge and reasoning. This is where Microsoft Foundry comes in. It’s the evolution of what used to be Azure AI Studio, then Azure AI Foundry, and now Microsoft Foundry. Foundry is where models are selected, grounded in enterprise content, evaluated, secured, and managed.

But architecturally, Foundry IQ handles the most difficult part of agent design: knowledge retrieval and grounding.

It connects agents to policy-controlled knowledge bases across Microsoft 365, cloud storage, data platforms, and internal repositories — all through a unified retrieval engine. Agents can perform iterative search, multi-source reasoning, and permission-aware grounding instead of relying on brittle manual retrieval stacks or static document embeddings.

Foundry IQ is what happens when AI understands contracts, policies, procedures, SLAs, regulatory constraints, and unstructured documents — and can reason across them safely. It’s the reasoning layer that makes AI context-aware instead of generic.

In modern architectures, this becomes the knowledge-access layer. It enables traceability, auditing, and reliable decision-making for autonomous workflows. Instead of an agent simply generating an answer, it can show where the information came from, respect access boundaries, and operate within governance constraints.

In short, Foundry IQ gives AI controlled, secure access to enterprise knowledge — and the ability to reason over it responsibly.

How the Three IQ Layers Work Together

If you want a simple picture in your head, imagine a three-layer stack. At the foundation is Fabric IQ, your structured data intelligence (data-context). In the middle is Foundry IQ, your reasoning and knowledge grounding layer (knowledge-context). On top is Work IQ, your understanding of how humans are actually operating day to day (user-context). Together, those layers create the conditions for meaningful enterprise agents.

Let’s make this concrete.

Suppose a company wants to build an AI agent to help manage supply chain delays. This is not a theoretical example. This is the kind of use case I hear from customers constantly.

Fabric IQ detects anomalies in delivery metrics. It sees that certain suppliers are trending late beyond historical norms. It notices that on-time delivery percentages are dipping in specific regions. It correlates delays with upstream bottlenecks. This is data-driven awareness.

Foundry IQ then grounds the agent in supplier contracts, SLAs, penalty clauses, and internal policies. It understands what the agreement actually says about late deliveries. It interprets escalation thresholds. It knows which suppliers have stricter terms and which ones allow flexibility. This is contextual reasoning.

Work IQ observes that operations teams are overloaded handling these exceptions manually. It sees long email chains, recurring “delay review” meetings, and individuals spending hours every week tracking updates from vendors. It identifies patterns of reactive work that are consuming capacity. This is workflow intelligence.

Now you introduce the Agent. The agent combines those three streams of intelligence. It recommends which delays need escalation based on contractual impact. It drafts communications to suppliers referencing the correct SLA language. It suggests internal reprioritization. It surfaces issues before they become crises. It becomes a force multiplier instead of just another dashboard.

Runtime, Governance, and Industrial Scale

But this is where responsible leaders pause and ask the right question. Who monitors this agent? Who ensures it’s accurate? Who ensures it stays within policy? And maybe the most important question: who is accountable when it makes a recommendation that affects the business?

That’s where Agent 365 comes in. Agent 365 is the runtime and governance layer that sits over everything. It monitors agent decisions, tracks accuracy over time, enforces compliance boundaries, and provides deep observability into how agents are behaving in the real world. It gives you visibility into what the agent is doing, why it is doing it, and whether it is staying within the guardrails you defined.

This isn’t just about logging activity. It’s about operational control. It’s about knowing when performance drifts, when policies change, when human override is required. It’s about giving leadership confidence that AI is not operating in the shadows but within a governed framework.

Without this layer, you have experiments — clever, promising, but fragile. With it, you have enterprise-grade systems that can scale responsibly.

And once this supply chain pattern works, Agent Factory allows you to scale it. Agent Factory is about industrializing agent creation. Instead of building one-off AI projects in isolation, you create repeatable blueprints that can extend to finance for invoice exceptions, HR for policy guidance, or operations for maintenance alerts. It’s the production line for enterprise agents.

Most companies today are experimenting with AI in pockets — a pilot here, a proof of concept there. It works, but it doesn’t scale. Agent Factory shifts you from experimentation to manufacturing. You define standard patterns for how agents are grounded in Fabric data, reason through Foundry, and are governed through Agent 365. You build templates, not just individual agents.

So when another department needs an agent, you don’t start from scratch. You extend the blueprint. That’s the difference between isolated AI projects and a true enterprise AI platform — designed once, governed consistently, deployed many times.

Now let’s untangle the product names, because this is where confusion often peaks.

Build and Orchestration Tools

To turn strategy into reality, you need build and orchestration tools. This is where Microsoft Foundry, Copilot Studio, and Copilot Studio Light come into play.

Microsoft Foundry is the engineering backbone for enterprise AI. It’s where foundation models (such as large language models like GPT-4-class systems), embedding models for retrieval and vector search, and task-specific AI systems like classifiers or anomaly detection models are selected, configured, grounded in enterprise data, evaluated for quality and safety, and prepared for production use. If you’re building serious AI solutions — the kind that must meet performance standards, compliance requirements, and security controls — this is where that work happens.

You can absolutely build agents in Microsoft Foundry. But you’re building them at the engineering layer. That means defining prompts, retrieval pipelines, tool calling, evaluation loops, safety filters, deployment endpoints — often in a pro-code environment. Foundry is not just a playground for models. It’s an engineering environment. It supports model evaluation, grounding techniques, prompt flows, fine-tuning scenarios, safety configurations, and deployment management. In other words, it’s where AI moves from experimentation to disciplined engineering.

Copilot Studio sits one layer above that. It’s the orchestration platform. This is where you design agents visually, connect them to business systems, define actions, build workflows, and determine how agents interact with users and applications. Copilot Studio is about behavior. It’s where you decide what the agent can do, what systems it can call, how it responds to certain triggers, and how it escalates when needed.

The clean distinction is this:
Foundry is pro-code, engineering-first agent construction.
Copilot Studio is low-code/no-code, workflow-first agent orchestration.

If Foundry is about building intelligence correctly, Copilot Studio is about putting that intelligence to work inside real business processes.

Copilot Studio Light — formerly known as Agent Builder — is the streamlined, no-code experience embedded directly inside Microsoft 365 Copilot. It allows business users to quickly create internal, knowledge-based agents grounded in SharePoint sites, Teams content, or organizational documents. It’s intentionally simple. It lowers the barrier to entry. But it’s important to understand the distinction: Copilot Studio Light is ideal for quick internal assistants and knowledge helpers. When you need advanced workflows, external system integrations, deeper customization, or structured governance controls, you step up to full Copilot Studio — and potentially integrate with Microsoft Foundry for more advanced model engineering.

When you step back, the architecture becomes clearer. Microsoft Fabric provides structured, governed enterprise data. Microsoft Foundry engineers and evaluates the intelligence layer — and can be used to build highly customized agents directly in a pro-code way. Copilot Studio is where you build and orchestrate business agents using a low-code/no-code approach, including their workflows, actions, and system integrations. Copilot Studio Light enables fast, internal, knowledge-based agents created directly inside Microsoft 365 Copilot. Together, they form a layered system rather than a collection of disconnected tools — pro-code where you need maximum control, and low/no-code where you need speed and broad adoption.

Here’s my slightly opinionated take. Microsoft isn’t just releasing features. It’s assembling an AI operating model for the enterprise. But when terminology evolves quickly — and it has — even experienced leaders can feel like they’re chasing moving targets. The risk is that we focus on memorizing product names instead of designing cohesive systems.

The real question isn’t whether you’ve tried Copilot Studio Light or deployed something in Microsoft Foundry. The real question is whether you are building isolated AI experiments or a layered, governed, repeatable AI system. Are your agents grounded properly? Are they orchestrated thoughtfully? Are they designed with scale in mind?

That’s the shift. Not from one product to another — but from experimentation to architecture.

Start simple. Pick one high-impact use case. Ground it in Fabric. Reason through Foundry. Understand human impact through Work IQ. Deploy responsibly with Copilot Studio. Govern with Agent 365. Then scale with Agent Factory.

I’ve been in this industry long enough to see naming waves come and go. The technology shifts. The labels evolve. But the pattern of success remains consistent. The companies that win don’t chase terminology. They understand architecture. They build deliberately. They scale responsibly.

If you’ve felt overwhelmed by the recent flood of AI terminology, you’re not alone. I was there too. But once you see the layered model underneath, the strategy becomes much less intimidating. And that’s when you can stop decoding names and start building intelligence.

Summing it up

Fabric IQ = Intelligence grounded in your structured business data
(Architectural label — not a SKU.)

Foundry IQ = Intelligence grounded in enterprise knowledge, policies, and reasoning
(Architectural label — not a product name.)

Work IQ = Intelligence about how work actually happens across meetings, messages, and collaboration patterns
(Architectural label describing direction.)

Microsoft Foundry = Where foundation models (LLMs, embedding models, task-specific models) are selected, grounded, evaluated, and engineered — and where highly customized agents can be built in a pro-code environment.

Copilot Studio = Where business agents are built and orchestrated using a low-code/no-code approach — defining workflows, actions, triggers, and system integrations.

Copilot Studio Light = Where quick, internal, knowledge-based agents are created directly inside Microsoft 365 Copilot for fast adoption and simple use cases.

Agent 365 = The runtime layer for monitoring, observability, compliance, and governance across deployed agents.

Agent Factory = The enterprise pattern for scaling and industrializing agent creation across departments.

The simple mental model:

Architectural layers create intelligence.
Engineering and orchestration tools build agents.
Agent 365 governs them.
Agent Factory scales them.

Pro-code where you need control.
Low-code/no-code where you need speed.