Microsoft uses the term Work IQ to describe the intelligence layer that Microsoft 365 Copilot sits on top of. Work IQ is not a separate product you buy — it is Microsoft's umbrella label for the data, context and skill layers that make Copilot personalized and tenant-aware.

Important framing: An earlier version of this article made claims that went beyond what Microsoft officially documents — particularly around end-user model choice and the availability state of the Work IQ API. We have rewritten the text and now distinguish explicitly between GA, preview and announced functionality. The canonical source remains Microsoft Learn.

The three layers of Work IQ

Microsoft frames Work IQ as the interplay of three layers:

1. 1Data — the content Copilot can access on a per-permission basis
2. 2Context — what Copilot knows about people, relationships, projects and work patterns
3. 3Skills and tools — specialized capabilities for concrete tasks

Layer 1: Data

The Work IQ data foundation includes:

• Microsoft 365 tenant data: SharePoint, OneDrive, Word, Excel, PowerPoint, Outlook, Teams — always permission-based, always honoring existing Sensitivity Labels and DLP policies.
• Copilot Connectors: Pre-built and custom connectors that bring third-party content into the index.
• Dataverse / Dynamics 365: Structured business data as an additional knowledge source. Microsoft has further integrations on the roadmap; exact availability per product and region is subject to change.

Layer 2: Context

Memory is the memory feature available in Microsoft 365 Copilot Chat with two facets:

• Explicit: Custom Instructions (e.g. "Always answer in active voice") and Saved Memories that users set themselves.
• Implicit: Copilot derives recurring preferences from chat history. Which signals feed into this — and when other activity (Outlook, Teams, Office) starts contributing — Microsoft is communicating in stages.

Semantic Index enables meaning-based retrieval across tenant content, not only keyword matching. Content from Copilot Connectors is also indexed; existing permissions remain authoritative.

Business Understanding is a semantic layer of glossaries and ontologies that helps Copilot interpret organization-specific terminology. Its depth depends on how consistently a tenant maintains SharePoint structures, term stores and sensitivity labels.

Layer 3: Skills and tools

Skills are specialized instructions for recurring tasks — for example scheduling a meeting, summarizing a briefing, interpreting a dashboard. Microsoft is building skills both into the first-party Copilot experiences and into Copilot Studio, where organizations build their own agents.

Model selection: what Microsoft actually offers today

This is where the previous version of this article overreached. As of today:

• Microsoft 365 Copilot uses a number of foundation models behind the scenes. Microsoft publicly discusses partnerships with, among others, OpenAI and Anthropic, and has announced that further models will be integrated into the platform.
• A direct model picker for end users is not available in every Copilot experience. It is being rolled out in specific surfaces (e.g. Copilot Chat / Researcher); in Word, Excel, Outlook or Teams there is usually no model selector at the prompt.
• What is available where changes frequently. The authoritative source is the matching entry in the Microsoft 365 Roadmap or Microsoft Learn.

We recommend not making multi-model arguments the central justification for Copilot investments. The value sits in the interplay of data, context, skills and governance — not in choosing a model name.

What is GA, preview or announced today

This table is intentionally a coarse orientation. For every individual feature: verify on Microsoft Learn before planning anything productive.

Security, permissions and compliance

Work IQ is not a new permission model — it uses the existing one:

• Permissions: Copilot sees only content the requesting person can access.
• Sensitivity Labels: Existing classification is inherited.
• DLP policies: Stay active.
• EU Data Boundary: Microsoft operates a dedicated data residency for the EU; details and exceptions are in Microsoft's EU Data Boundary documentation.
• Tenant boundary: Tenant data is not used to train third-party foundation models.

For Swiss organizations with FADP, FINMA or sector-specific obligations, the implication is straightforward: the governance homework (permissions, labels, DLP, external access) is the same with or without Copilot. The cleaner that homework is done, the more value Copilot can deliver.

What you can realistically do today

1. 1Tighten tenant hygiene: refresh SharePoint permissions, external access, sensitivity labels and DLP policies — that is the highest-leverage step.
2. 2Start with a few well-defined use cases: measure three scenarios properly in a pilot rather than rolling out ten in parallel.
3. 3Use Memory deliberately: align Custom Instructions centrally instead of letting every person curate their own Saved Memories ad hoc.
4. 4Track roadmap updates: review Microsoft 365 Roadmap and Microsoft Learn regularly instead of relying on marketing decks.
5. 5Measure outcomes: before/after measurement on concrete tasks — not just generic adoption KPIs.

How CNEXT can help

As a Microsoft Partner, we accompany Swiss organizations through an honest Microsoft 365 Copilot rollout:

• Readiness assessment: tenant hygiene, permissions, labels and DLP.
• Use-case workshop: realistic scenarios for your industry and your data.
• Pilot support: measurable rollout with clear success criteria.
• Training & enablement: end users, power users and IT admins — based on what is actually available.

Work IQ as a concept is useful because it raises the right questions: Where is our data? How clean is our context? Which skills do we need? The answers are not in a marketing term, but in the tenant reality of each individual organization.