If you followed any of the noise from Microsoft Build 2026, you would have heard the term Microsoft IQ a great deal. It was the unifying idea of the event. It is also one of those announcements that is easy to nod along to and hard to actually explain. This guide fixes that - no jargon, no acronym soup, just what it is and why it should be on your radar.

The problem Microsoft IQ is built to solve

AI agents are clever about language and almost completely ignorant about your organisation. A capable model knows how to write an email; it has no idea who your customers are, what your processes look like, or what your internal terms mean. Ask it to act on your behalf and it has to guess.

You cannot fix that by making the model bigger. You fix it by giving the model context - grounding it in your organisation's real data and knowledge. That grounding is what Microsoft IQ provides. Microsoft describes it as a "context layer": something that sits between your data and your agents, translating raw organisational information into something an agent can reason about reliably. It went generally available on 2 June 2026, and works across Microsoft's agent surfaces including GitHub Copilot, Microsoft Foundry and Copilot Studio.

The simplest way to think about it: the model is the brain; Microsoft IQ is the memory and context that brain needs to be useful inside your business.

The four layers, in plain English

Microsoft IQ is not a single product. It is an umbrella over four kinds of context, each answering a different question an agent needs answered.

Work IQ - "how work happens here"

Work IQ draws on the signals inside Microsoft 365: the people, documents, emails, meetings and chats, and the relationships between them. It lets an agent understand that a particular person is on a particular team, that a relevant document lives in a particular place, and who works on what. Crucially, it stays inside your Microsoft 365 trust boundary and respects existing permissions.

Fabric IQ - "what the business means"

Fabric IQ is the shared layer of business meaning over your structured data. It is where your data's semantics live - the agreed definitions of your core business concepts and how they relate. If Work IQ is about how work happens, Fabric IQ is about how the business is defined. It builds on your data foundation in OneLake and on the semantic models your analytics teams already maintain.

Foundry IQ - "what we already know"

Foundry IQ is about knowledge and retrieval: letting agents discover and reuse knowledge rather than starting from scratch each time. It unifies different knowledge sources behind a single retrieval endpoint, so an agent can pull the right information at the right moment. For organisations sitting on large volumes of documents and unstructured knowledge, this is the layer that turns that content into something agents can actually use.

Web IQ - "what's true in the world right now"

The newest of the four, Web IQ adds live, real-time context from the open web - so an agent's view is not frozen at its training cut-off. It handles the grounding in current external information that internal data alone cannot provide.

Together, these four give an agent a far richer picture: how your organisation works, what your data means, what you already know, and what is happening in the world.

Why this matters for your organisation

The headline implication is strategic, not technical. Context is becoming the real differentiator in enterprise AI. Two organisations using the same model will get very different results depending on how well their agents are grounded in their own context. As Microsoft frames it, organisations that build on this context layer early will accumulate advantage with every deployment; those that keep deploying isolated, ungrounded agents will keep rebuilding the same foundations.

For regulated and data-sensitive organisations there is a second implication: a context layer is only as trustworthy as the data and permissions beneath it. An agent grounded in messy, ungoverned data is grounded in the wrong thing. Which leads to the uncomfortable but important point.

The catch: context only works if your foundation is ready

Microsoft IQ is a powerful way to deliver context to agents. It does not create good context out of bad data. If your data is fragmented or duplicated, your business terms are undefined, or your governance is loose, the context layer will faithfully serve those problems to every agent you run.

This is why we tell clients that the context layer is a reason to get the fundamentals right, not a way to skip them. The organisations that will get the most from Microsoft IQ are the ones whose data is unified, matched and governed, and whose business definitions are agreed - in other words, the ones that are genuinely agent-ready.

Where to start

If Microsoft IQ is on your roadmap, the most useful first move is not a platform decision - it is an honest look at the foundation underneath it. Our work with organisations adopting the context layer almost always starts with two questions: is your data ready to be grounded on, and is your knowledge in a state an agent can use?

To find out, start with our Agent-Ready Data checklist, or talk to us about what getting ready for the context layer looks like for your organisation.

‍