Microsoft's new IQ layer is less about another chatbot and more about owning the context layer that enterprise agents need to do useful work.
Microsoft used Build 2026 to make a very clear point: the next fight in enterprise AI will not be won by the company with the flashiest demo. It will be won by the company that can connect agents to the messy reality of business data, permissions, workflows, and live information without forcing every developer to rebuild the same plumbing from scratch.
That is what Microsoft IQ is meant to be. Announced at Build on June 2, the platform brings together Work IQ, Fabric IQ, Foundry IQ, and Web IQ into a shared intelligence layer for agents across GitHub Copilot, Microsoft Foundry, and Copilot Studio. According to Microsoft's Build 2026 post, Microsoft IQ is generally available now, while Work IQ APIs are scheduled for general availability on June 16.
The timing matters. Companies have spent the past two years experimenting with AI assistants that can draft emails, summarize documents, or answer support questions. Useful, yes. Transformative, not always. The harder problem is getting an agent to understand the business well enough to make a decision, take action, and still respect the company's rules about access, identity, and governance.
Enterprise agents fail for predictable reasons. They do not know which version of a document is authoritative. They cannot tell whether revenue means booked revenue, recognized revenue, or pipeline. They lose the thread when work moves between Teams, Outlook, SharePoint, GitHub, Salesforce, ServiceNow, and a dozen internal systems. So employees keep checking the work manually, which defeats the point.
Microsoft's answer is to make context reusable. Work IQ is built around how work happens inside Microsoft 365, including people, meetings, emails, files, roles, and collaboration patterns. Fabric IQ is meant to give agents a shared semantic foundation over structured business data, so different agents reason from the same definitions. Foundry IQ turns fragmented knowledge into governed retrieval for agent applications. Web IQ adds fresh external information, including web pages, news, images, and videos, using Microsoft's search infrastructure.
That sounds abstract until you think about the everyday use case. A sales agent preparing an account brief needs internal relationship history, the latest news about the prospect, approved product language, current pricing rules, and perhaps a support issue that is sitting in another system. If each of those pieces arrives through a separate brittle integration, the agent becomes expensive to maintain. If the agent can draw from one governed context layer, the economics start to look different.
Microsoft is also pushing the token and latency argument. In a Microsoft 365 blog post, Charles Lamanna's team said Work IQ APIs are designed to return agent-ready context rather than raw content that another orchestration layer must interpret. The same post said the APIs collapse tool operations into 10 generic tools through Model Context Protocol and use fewer tokens by moving more processing into the Work IQ runtime. For businesses running agents at scale, that is not a small detail. Cost and speed decide whether a prototype becomes a workflow.
Hosted agents bring the runtime fight into Foundry
The second piece is hosted agents in Foundry Agent Service, expected to reach general availability by early July 2026. Microsoft's Foundry blog says each hosted-agent session gets its own sandbox with dedicated compute, memory, and filesystem access, while supporting frameworks such as Microsoft Agent Framework, GitHub Copilot SDK, LangGraph, and other SDKs without rewrites.
This is where Microsoft is trying to do for agents what cloud platforms did for applications. Developers can build in GitHub, deploy in Foundry, publish into Teams or Microsoft 365 Copilot, and operate with tracing, evaluation, memory, and governance in one environment. That is a strong pitch for the enterprise buyer who wants agents, but does not want hundreds of disconnected experiments running across departments.
It also creates a sharper question for startups. Building on Foundry could give young companies access to Microsoft's enterprise distribution, identity model, data connectors, and customer trust. That is a serious advantage if the product is meant to live inside large companies rather than around them.
But dependence comes with limits. If a startup's main value is orchestration, retrieval, memory, or agent hosting, Microsoft is moving directly into that lane. The opportunity shifts toward industry-specific workflows, proprietary data, vertical expertise, compliance depth, and user experience. In other words, startups may need to build higher up the stack, where Microsoft provides the pipes but not the finished business outcome.
This is also why Amazon and Google cannot treat Build 2026 as just another developer event. AWS Bedrock Agents and Google Vertex AI Agent Builder already give enterprises ways to create agentic systems, but Microsoft has a distribution advantage inside everyday work. Teams, Outlook, SharePoint, GitHub, Excel, Power Platform, and Microsoft 365 Copilot are not side channels. They are where many employees already spend the day.
The practical test will come after the announcements settle. Enterprises will want to see whether Microsoft IQ can reduce custom integration work, improve answer quality, and keep permissions intact when agents move from search to action. Developers will want clear pricing, reliable APIs, and enough flexibility to avoid being trapped in one architecture.
Microsoft has not solved the enterprise agent problem just by naming an intelligence layer. But it has made the battlefield clearer. The next stage of AI adoption will be about context, runtime, governance, and distribution. Watch how quickly companies move agents from pilots into real workflows, because that will tell us whether Microsoft IQ is a platform announcement or the beginning of a new enterprise software stack.
Also read: DeepSeek exposes the price pressure building inside enterprise AI • DeepSeek is making AI cost discipline impossible to ignore • South Korea puts a Naver veteran at the center of its AI push
