Chat, Context, Tools, Workspaces: The Four Domains of the Work IQ APIs
The new Work IQ APIs look like a developer release. In practice, they show how Microsoft wants to scale enterprise agents: with a runtime that supplies context, tools and persistent state inside the Microsoft 365 tenant. The API surface covers four domains: Chat, Context, Tools and Workspaces. Billing runs on Copilot Credits with spending limits for admins. In this article, I walk through the four domains and explain what they mean for developers, partners and IT leaders.
Microsoft's new Work IQ APIs are easy to read as a pure developer release. They also show how Microsoft plans to scale enterprise AI. Agents that retrieve context, take actions and work on a task over several steps need infrastructure, not only a prompt box.
What Microsoft announced
Microsoft announced general availability of the Work IQ APIs as the interface through which agents interact with Microsoft 365 data and apps. Work IQ continuously processes signals from email, calendar, meetings, chats, files, people, collaboration patterns and line-of-business systems. From these signals, it builds an understanding of who works with whom, which artifacts are relevant and how work progresses. Raw APIs for mailboxes and files deliver data. Work IQ delivers interpreted context.
Four domains
Chat The Chat API returns the answer Microsoft 365 Copilot would give a user, including citations. Developers embed grounded Copilot answers in their own applications without building a reasoning stack.
Context The Context API returns the context Copilot would use for a query, formatted for agents rather than human readers. Classic retrieval delivers documents and snippets, and the orchestration layer then spends tokens sorting them. Prepared context reduces latency and token usage and keeps results consistent across workflows.
Tools Work IQ reduces tool calling to a small set of generic tools. Anyone who has built agents knows the problem: the more specific actions a model can choose from, the more often it picks the wrong one. A stable, compact tool surface makes agents faster and more reliable.
Workspaces Agents store intermediate state, files, memory and temporary outputs in digital workspaces inside the tenant boundary. Long-running tasks need this, because they collect information over time and hand results back to users for review. Microsoft links these workspaces to its own agents Copilot Cowork and Microsoft Scout.
Governance and cost
According to Microsoft, data, context and insights stay within the Microsoft 365 tenant, and actions remain auditable. Organizations need to know which data an agent could access, which actions it took under which identity, what it stored and what it cost.
Billing uses Copilot Credits. There is no separate Work IQ subscription and no per-user license. Tool API calls cost a fixed amount, while Chat and Context usage varies with scenario complexity. Admins can review usage and set spending limits. AI cost management thus resembles cloud operations: organizations monitor and tune consumption continuously instead of deciding once at procurement.
Work IQ in Microsoft's stack
Work IQ completes a series of announcements: agentic actions in Word, Excel and PowerPoint, business apps inside Copilot, long-running agents such as Cowork and Scout, model choice across providers and orchestration in Copilot Studio. Work IQ forms the intelligence and runtime layer beneath these products. Microsoft positions Microsoft 365 as the environment in which enterprise agents are grounded, governed and executed.
What this means for developers and IT leaders
Developers no longer have to build retrieval pipelines, access management, tool orchestration, memory handling and cost tracking themselves. The complexity does not disappear, but it moves. Teams can focus on domain logic and workflows, where the business value lies.
For CIOs, the question is whether the organization wants a common foundation for agents. Without one, disconnected copilots, duplicated retrieval systems, fragmented governance and poor cost transparency emerge. A shared layer prevents this and extends existing Microsoft 365 investments to agent infrastructure.
Enterprise AI is often framed as a race for the best model. Long-term advantage may instead lie with whoever controls the layer that connects models, business context, tools, memory, governance and billing. Work IQ is Microsoft's bid for this layer.