Work IQ API: how Microsoft opens the intelligence layer behind Copilot to custom AI agents on June 16

If you have used Copilot in Microsoft 365 and ever been surprised by how it retrieves relevant emails, colleagues and meeting notes, then you have seen Work IQ in action without knowing it. Work IQ is the intelligence layer that gives Microsoft 365 Copilot semantic understanding of your organisation. On 2 June 2026, Microsoft announced that this layer becomes available via a public API on 16 June to anyone who wants to build their own AI agents on Microsoft 365 data.

That is a fundamentally different step than earlier extensions to the Microsoft 365 platform. Previously you could create connections via Microsoft Graph to access individual data streams. What is now opening is not a raw data connection, but a pre-processed, semantically enriched intelligence layer — the same one Copilot uses every day. This article explains what Work IQ does, what exactly becomes available on 16 June, and what the consequences are for organisations that are serious about building their own AI agents.

What Work IQ does: from raw data to business understanding

Microsoft 365 produces enormous amounts of data every day: emails, meeting transcripts, chat messages, documents, tasks, calendar events. Most systems approach that data as discrete objects. Work IQ does something different: it processes all that information continuously and builds a semantic model of your organisation. Who collaborates with whom? Which projects are active? What are the communication patterns between departments? What skills does an employee have?

The result is not a database but a context structure. Work IQ uses a semantic index with low latency that not only searches by keyword but understands the meaning behind a question. When Copilot is asked for a status summary of a project, it does not only retrieve files containing that project name. It understands which people are involved in the project, what was discussed in meetings over recent weeks, which tasks are open, and which documents have been recently updated.

The three layers of Work IQ

Work IQ is built from three closely integrated layers. The first layer is data: everything in your Microsoft 365 tenant, from SharePoint files to Teams messages and Exchange calendars. That data always stays within the tenant boundary — Work IQ does not move company data to a central Microsoft pool.

The second layer is context. Work IQ adds meaning to the raw data through the semantic index and through personal memory. It remembers how people in your organisation work, which projects have priority, who the decision-makers are for which file, and how quickly someone typically responds to a particular type of message. That context makes responses not just correct but also relevant to the person asking the question.

The third layer is skills and tools. Work IQ provides agents with ready-to-use actions they can perform within the Microsoft 365 ecosystem: send an email, schedule a meeting, upload a document, create a task in Planner, send a Teams message. An agent working through Work IQ does not need to programme those actions itself — it inherits them as part of the connection.

What opens on 16 June: the Work IQ API

From 16 June 2026, three access forms for the Work IQ API are generally available: a REST API, a remote MCP server, and an A2A interface (agent-to-agent). The REST API is the most direct way for developers to integrate Work IQ into an existing application or a custom agent built outside the Microsoft ecosystem. The remote MCP server connects to the Model Context Protocol, an open standard that an increasing number of AI frameworks support as a way to provide agents with external context. The A2A interface enables Microsoft 365 agents to collaborate with agents from other platforms through a structured protocol.

The most important difference from the previous situation is what an agent builder no longer needs to do. Previously, a developer wanting to build a business agent on Microsoft 365 data had to work connector by connector: a separate Graph connection for email, another for calendar, a third for meeting metadata, and so on. Each connection required its own authentication, data modelling and contextualisation. Via the Work IQ API, a single connection is sufficient to gain access to that integrated context — including semantic understanding, personal memory, and ready-to-use actions.

Workspaces as the foundation of agents

An important new concept introduced with the Work IQ API is digital workspaces. A workspace is a bounded environment within your Microsoft 365 tenant where an agent maintains its intermediate results, files, memories, and progress while executing a complex task. The workspace does not exist outside the tenant boundary — all data that an agent collects and produces remains within your own environment.

This is a direct response to one of the biggest concerns around AI agents in enterprise environments: where does the data go when an agent executes a multi-day task? With Work IQ workspaces the answer is: nowhere. The agent works in your tenant, stores in your tenant, and the data is addressable through your existing governance and compliance policies, including Microsoft Purview.

Costs: Copilot Credits as the billing model

The Work IQ API does not operate through a separate subscription or per-user licence. Usage is billed through Copilot Credits — the unified consumption currency that Microsoft also uses for Copilot Studio flows and other AI services within the Microsoft 365 ecosystem. Organisations that already have a Copilot Credits budget set up for Copilot Studio do not need to sign a separate contract for the Work IQ API.

Microsoft has added new management capabilities in the Microsoft 365 Admin Center: administrators can set spending limits per agent or per application, monitor usage, and configure alerts when a specific consumption threshold is reached. That makes the model transparent and controllable, even when multiple agents use the API concurrently.

What this means for your IT strategy

The opening of the Work IQ API marks a shift in how Microsoft 365 functions as a platform. Until now the primary access point for AI was Microsoft Copilot itself — an interface built by Microsoft on an intelligence layer managed by Microsoft. From 16 June, organisations can access that intelligence layer for their own applications. That opens a range of concrete scenarios that were previously impossible without extensive custom integrations.

A sales team can build an agent that combines customer communication in Exchange and Teams with CRM data and automatically generates a complete view of a customer's status before a sales call. An HR department can deploy an agent that guides onboarding processes with access to all relevant documents, meeting history, and task lists for the new employee. An IT team can develop an agent that correlates incident reports with recent changes and meetings to identify the likely cause of a problem.

What all these scenarios have in common: the agent understands the business context through Work IQ, acts autonomously but within the existing governance structure, and stores its results within the tenant boundary. The technical complexity of building this has dropped significantly now that it no longer starts from individual Graph connections but from a ready-made context layer.

What you can do this week

Four steps for organisations that want to seriously explore the Work IQ API. First: read the documentation on Microsoft Learn about the Work IQ API endpoints and the MCP server option. The documentation provides a concrete picture of what is and is not exposed through the API. Second: determine whether your organisation already has a Copilot Credits budget. If you use Copilot Studio, that is likely already the case. If not, ask your Microsoft partner about the options around consumption subscriptions.

Third: inventory which internal processes would benefit from an agent that has access to the full context of your Microsoft 365 environment. Think about processes where employees currently combine multiple sources themselves — email, meetings, documents, tasks — to determine a status or make a decision. Those are the first candidates for automation via Work IQ-based agents.

Fourth: involve your IT administrator in the governance question. Work IQ agents operate within your tenant but have contextual access to a broad spectrum of business data. Determine in advance which data types are accessible to which type of agent, and how you enforce that policy through the spending limits and access controls in the Admin Center. Want support exploring the Work IQ API, building a first agent on your Microsoft 365 environment, or drafting a governance framework for business agents? Contact Zarioh for a no-obligation conversation.