The Context Layer Goes Live
Microsoft's Work IQ APIs reached general availability on June 16, billed as the primary way for AI agents to interact with Microsoft 365 data and applications. Work IQ is the intelligence layer that builds a semantic understanding of a business by continuously processing email, calendar, meetings, chats, files, people, collaboration patterns, and line-of-business systems. In plain terms, Microsoft is exposing a structured, queryable model of how an organization actually works, and inviting AI agents to build on it.
We regard this as one of the more strategically consequential enterprise AI moves of the year, even if it lacks the drama of a new model launch. The persistent challenge with enterprise agents has been context: a capable model that knows nothing about your meetings, documents, and relationships is of limited use. Work IQ is Microsoft's answer, and because so much of the corporate world already runs on Microsoft 365, the company is positioned to own the context layer that agents depend on. Whoever controls that layer holds enormous leverage.
Three APIs and a Workspace
The GA release ships three core surfaces. The Chat API provides programmatic access to Copilot responses with citations. The Context API delivers agent-ready context and aggregates sources. The Tools API exposes Microsoft 365 entities through simplified verbs and resource paths. Alongside them, Workspaces offer secure intermediate state storage within tenant boundaries, and the release includes Model Context Protocol integration, support for agent-to-agent communication, a redesigned remote MCP server, and a REST API.
The architecture reveals Microsoft's intent to be the platform, not just a participant. By supporting open standards like the Model Context Protocol and agent-to-agent patterns, Microsoft is signaling that it wants Work IQ to be the foundation other vendors' agents build on too, not only its own Copilot. We find that posture telling. The company has learned that the durable position in any platform war is to be the substrate everyone integrates with. Embracing interoperability while owning the underlying context is how Microsoft intends to remain indispensable as the agent ecosystem fragments across vendors.
Consumption Pricing Arrives for Agents
The commercial model deserves attention because it changes how enterprises will budget for AI. Pricing is consumption-based through Copilot Credits, a unified currency that also covers Copilot Studio, with a fixed component for Tools and a variable component for Chat and Context. There is no separate Work IQ SKU and no per-user license. This is a meaningful departure from the per-seat licensing that has defined Microsoft's productivity business for decades.
We see real implications in that shift. Per-seat pricing is predictable; consumption pricing is not, and it ties cost directly to how heavily agents actually use the context layer. An organization that deploys agents aggressively could see costs scale in ways that are hard to forecast in advance. The move to credits acknowledges that agent workloads are inherently variable, but it also transfers planning risk to the customer. Finance and IT leaders will need new disciplines to manage AI spend that behaves more like cloud compute than like a fixed software subscription.
Governance Built for the Real Concern
Microsoft clearly understands that the barrier to enterprise agent adoption is governance, not capability, and the release is built accordingly. A new cost-management dashboard in the Microsoft 365 admin center lets IT administrators configure billing, set spending limits per tenant, per group, and per user, and monitor credit usage. Crucially, the entire model is permission-aware: agents see only what the underlying identity is allowed to see, so existing access controls extend to agent activity rather than being bypassed by it.
This is the right emphasis. The nightmare scenario for any CISO is an AI agent that helpfully surfaces information a user was never permitted to access, turning a productivity feature into a data-leak engine. By binding agent access to identity and existing permissions, and giving administrators granular spend and usage controls, Microsoft is addressing the objections that stall deployments. We have consistently argued that the winners in enterprise AI will be those who solve governance, and Work IQ's design suggests Microsoft has internalized that lesson.
The Platform Play Comes Into Focus
Stepping back, Work IQ is the connective tissue for Microsoft's broader agent ambitions, including its Scout Autopilot agent and the wider Copilot ecosystem, all led by Charles Lamanna, the executive vice president for Copilot, agents, and platform. The strategy is coherent: own the productivity suite, expose its accumulated context through governed APIs, price it as consumption, and let an ecosystem of agents, Microsoft's and others', build on top. Each layer reinforces the others.
For enterprise technology leaders, the practical takeaway is that the foundation for production agents on Microsoft 365 is now in place, and the decision to build on it is no longer hypothetical. The questions shift from whether the platform exists to how to govern its cost and access. We expect Work IQ to accelerate enterprise agent deployment precisely because it removes the two hardest obstacles, context and permission control, at once. The risk worth naming is dependence: building your agent strategy on Microsoft's context layer is convenient and powerful, and it deepens a reliance that will be difficult to unwind later.
