Microsoft opens Work IQ APIs for enterprise agents

Microsoft says the Work IQ APIs are generally available on June 16. The announcement, part of its Build 2026 news, describes Work IQ as a workplace intelligence layer for software agents: a way to give them programmatic access to organizational context from Microsoft 365, internal systems and external sources. The central fact is narrow but important. Microsoft is not only offering models and agent tools; it is exposing a context layer that can help agents understand how work actually happens inside a company.

Work IQ sits inside Microsoft IQ, which Microsoft says is available across GitHub Copilot, Microsoft Foundry and Copilot Studio. The idea is easy to state and hard to execute. A workplace agent cannot be very useful if it has no structured understanding of people, emails, documents, meetings, business systems and the relationships among them. The Work IQ APIs are meant to make that context available to developers, so an agent can prepare a meeting, retrieve a decision, understand a workflow or take the next step without forcing the user to restate everything from scratch.

The shift matters because enterprise agents are becoming a context and governance problem as much as a model problem. Giving an AI system access to work data can make routine tasks faster, but it also raises hard questions about permissions, auditability, security and boundaries. Microsoft frames Work IQ alongside Fabric IQ, which provides a semantic foundation over structured business data, and Foundry IQ, which helps plan retrieval across enterprise knowledge and the live web. In practice, that points to a stack where an agent is connected to managed sources rather than left to improvise from a prompt window.

This is a more mature phase of generative AI in organizations. Early adoption often meant chat interfaces, manual file uploads and one-off prompts. The next layer tries to connect AI to ordinary work flows through APIs, policy controls and data graphs. That could make agents more useful, but it also makes them more dependent on the quality of the company’s data, the accuracy of permissions and the discipline of internal knowledge management. A weak document graph or unclear access model will not become reliable simply because an agent can call it.

The useful question is therefore practical rather than promotional. Can an agent work with the right context, at the right time, while avoiding information it should not see? Work IQ is Microsoft’s answer to that problem, and its general availability gives developers a more concrete surface for testing whether enterprise agents can move from demos to governed daily use.