Key takeaways
- “Microsoft IQ” is the name for the intelligence layer that grounds Copilot and AI agents in a shared understanding of your organisation.
- It brings together four capabilities: Work IQ (how people work), Fabric IQ (what the business is and does), Foundry IQ (what the organisation knows), and Web IQ (the outside world).
- Together they supply the three contexts AI needs to be reliable (user context, data context and knowledge context) plus fresh web context.
- Their real purpose is to reduce guesswork and hallucination, so AI reasons from your actual business rather than generic assumptions.
- Crucially, they respect existing identity and data-protection controls, so grounding never means bypassing security.
Why "IQ" suddenly matters
Anyone following Microsoft’s AI announcements over the past year will have noticed a new family of names appearing: Work IQ, Fabric IQ, Foundry IQ and Web IQ. Introduced as a coherent set around Microsoft’s Ignite 2025 conference, they can seem like just another set of product names to decipher. Behind them, however, is a useful idea that helps explain why some organisations get reliable, grounded results from AI while others struggle with generic or unreliable responses.
At the heart of the idea is this: models alone are not enough. A powerful AI model knows a great deal about the world in general, but nothing about your organisation in particular. It can write a polished project plan, yet it has no idea which projects you’re actually running, who is accountable for them, or what your last board paper said. To be genuinely useful at work, AI needs context. It needs to understand how your people work, what your business does, and what your organisation knows. Providing that context, under the right governance, is exactly what the IQ layers are designed to do.
Collectively, Microsoft refers to this as Microsoft IQ: an enterprise intelligence layer that gives Copilot and AI agents a shared, continuously updated understanding of your organisation. Rather than every AI tool having to work out how to connect to your business independently, the IQ layers provide a common foundation. It’s helpful to think of them as intelligence layers and capabilities (some still evolving or in preview) rather than standalone products you buy off the shelf. Some work quietly behind the scenes in Copilot, while others come into play when you build AI agents of your own.
The three contexts AI needs
One helpful way to think about the IQ family is through the different kinds of context AI needs to work effectively, plus a fourth that connects it to the outside world.
Picture an experienced new starter joining your team. To be effective, they need to learn how the organisation works day to day, understand what the business does and how success is measured, absorb the knowledge captured in policies, files and past projects, and keep an eye on what’s happening beyond the organisation. The IQ layers give AI those same four kinds of awareness.
Work IQ: How People Work (User Context)
Work IQ is the layer closest to everyday users. It draws on the everyday signals flowing through Microsoft 365, including documents, emails, Teams chats, meetings, calendars and the relationships between people, to build an understanding of how work actually happens in your organisation: who works with whom, which projects are active, and how work typically flows between teams.
Microsoft describes it as having three parts: the data (the signals from your Microsoft 365 activity), the memory (a model of your preferences, habits and your organisation’s real working structure, beyond the formal org chart), and the inference that connects the two to anticipate the next useful action.
The result is a more personalised experience. Ask Copilot to “summarise where we got to with the Henderson grant” and Work IQ is what helps it find the right conversation, the right people and the right documents, rather than asking you to spell out every detail. Importantly, Work IQ respects existing permissions, sensitivity labels and compliance controls, so it only surfaces information the person asking is already entitled to access.
Fabric IQ: What the Business Is and Does (Data Context)
Fabric IQ adds business meaning to your data. Rather than asking AI to reason over raw tables, columns and timestamps, Fabric IQ sits over your data estate and provides a semantic layer and an ontology: a structured model of your business entities, such as customers, services, cases and products, together with the relationships, rules and metrics that connect them.
The benefit is a more consistent understanding of your data. With a shared semantic model, an AI agent and a Power BI report can use the same definition of “active client” or “overdue”, helping avoid the inconsistencies that often arise when different systems define the same thing differently. Consider how often a single term means different things across an organisation. A community organisation might count an “active member” one way in its membership database, another way in finance, and a third way in a spreadsheet maintained by the events team. When AI reads those sources independently, it inherits that ambiguity. Define the entity once in Fabric IQ, and every report and agent can work from the same agreed meaning.
Fabric IQ also includes a graph engine for reasoning across connected entities and supports data and operations agents that can monitor the live state of the business and respond to it. In practice, it helps AI reason about your organisation using shared business definitions instead of disconnected data.
Foundry IQ: What the Organisation Knows (Knowledge Context)
Foundry IQ helps solve one of the biggest challenges in building reliable AI: getting the right knowledge to the right place at the right time. It turns your unstructured knowledge, such as policies, manuals, contracts, FAQs and past work, into reusable, governed knowledge bases that multiple agents and applications can draw on through a single connection.
This matters because, without it, every AI project ends up rebuilding the same foundations: connecting to content sources, indexing documents and managing permissions. Foundry IQ provides those capabilities as a shared service, bringing together knowledge from Microsoft 365, SharePoint, OneLake, the web and other sources into a single grounded endpoint.
Picture a small membership organisation that wants one agent to support staff and another to help members directly. With Foundry IQ, both agents point to the same governed knowledge base. When a policy document changes, every agent benefits from the update automatically instead of relying on separate copies that gradually drift out of date.
Just as importantly, Foundry IQ is built on Microsoft’s existing governance capabilities. It respects Microsoft Entra identity, user permissions, and the classifications and sensitivity labels managed through Microsoft Purview throughout the retrieval process. This allows agents to cite sources, respect permissions and retrieve information from trusted knowledge rather than relying on general knowledge alone, helping reduce inaccurate or unsupported responses.
Web IQ: The World Outside (Web Context)
Completing the picture, Web IQ provides AI systems and agents with fresh, real-world information from across the web, allowing grounded answers to combine current external information with your organisation’s own context. Plenty of useful work depends on information that sits outside your organisation: a funding body’s updated reporting requirements, a supplier’s published service status, or current public reference data. Web IQ allows an agent to incorporate that information while keeping the response grounded in your own knowledge and data.
How They Fit Together
Each layer is valuable on its own. They become much more useful when they work together, because most business questions draw on more than one kind of context.
Imagine an organisation managing recurring supplier delays. Fabric IQ (data context) identifies the pattern, showing which suppliers are slipping and where on-time performance is declining. Foundry IQ (knowledge context) retrieves the relevant supplier contracts and service-level agreements, so the obligations around those delays are clear. Work IQ (user context) recognises that the operations team is spending significant time chasing suppliers through emails and meetings.
The result is an AI agent that can do more than produce another dashboard. It understands where and why the problem is occurring, what the contract says about it, and how the team currently manages it. From there, it can recommend which delays should be escalated, draft an email referencing the relevant contract clause, and suggest improvements to the existing process.
The same pattern applies in a non-profit. A case worker preparing for a client review receives the client’s status and history from Fabric IQ, the relevant policies and eligibility rules from Foundry IQ, and the latest notes and key colleagues from Work IQ. One question produces a grounded, properly sourced briefing instead of a generic response that needs to be checked and corrected.
Taken together, these layers give AI a richer understanding of how your organisation works, what it knows and what is happening across the business. The same idea underpins our companion articles on the 365 Architechs AI Framework, where data sits at the foundation, and on building a second brain for the organisation.
| Layer | Provides | Helps AI Understand |
| Work IQ | User context | How your people work |
| Fabric IQ | Data context | What your business is and how it operates |
| Foundry IQ | Knowledge context | What your organisation knows (documents, policies) |
| Web IQ | Web context | What’s happening beyond your organisation |
How to Get Started
You don’t need to adopt the IQ layers all at once. In fact, it’s often more effective to start with a single, well-defined problem where better grounding would make a clear difference: a frequently asked policy question, a recurring report or a routine piece of correspondence. Working through one real use case usually teaches you far more than trying to tackle everything at once.
Before you build anything, spend some time on the foundations the IQ layers rely on. Foundry IQ and Work IQ can only work with the permissions and sensitivity labels that already exist, so the quality of your results depends on the quality of your underlying information. Review access to the sites and files your agent will use, apply sensitivity labels to genuinely confidential content, and make sure the documents you want AI to rely on are current and trustworthy.
For data-driven use cases, agree on the handful of business definitions and metrics that matter most. That’s the shared meaning Fabric IQ is designed to provide. From there, build one grounded agent, test it using real questions from your users, and expand once you’re confident in the results. It’s the same practical approach we take with clients: start with a focused use case, build on solid foundations and grow from there.
Why This Matters for SMEs and Not-for-Profit Organisations
At first glance, these capabilities can sound like something designed only for large enterprises. In practice, they can be just as valuable for smaller organisations.
One reason is trust. For organisations with limited time and resources, AI needs to be dependable enough to support real work. Grounding AI in your own data, knowledge and ways of working helps produce responses that are more relevant, consistent and reliable. Reducing inaccurate or unsupported answers is especially important when AI is being used to assist staff, members or clients.
The second reason is accessibility. Capabilities such as Foundry IQ mean smaller organisations don’t have to build every part of an AI solution themselves. Much of the work involved in connecting, governing and retrieving knowledge becomes part of the Microsoft platform, making it much easier to build useful agents without assembling a large development team.
Governance remains an important part of the picture. Because these layers respect existing Microsoft Entra identities and Microsoft Purview protections, grounding AI doesn’t require weakening your security. It does, however, depend on your underlying permissions and data classifications being in good shape. If permissions are too broad or sensitive content is incorrectly classified, AI will expose those issues more quickly simply because it can find information much faster than people can. That’s another reason to invest in good information governance before expanding your AI capabilities.
The Bottom Line
The Microsoft IQ family, including Work IQ, Fabric IQ, Foundry IQ and Web IQ, brings together the different kinds of context AI needs to work effectively: how your people work, what your business does, what your organisation knows and what’s happening beyond your organisation. Together, these intelligence layers help AI produce responses that are grounded in your organisation rather than relying on general knowledge alone.
For SMEs, corporates, and non-profits, that means AI can become a practical tool for everyday work, provided it’s built on good data, trusted knowledge and sound governance.
At 365 Architechs, we help organisations build those foundations so AI is grounded in the way your organisation actually works, supported by the right data, knowledge and governance from the outset.
Want AI that’s grounded in your business, not generic assumptions? We’d be happy to help you build the right foundations.
