Microsoft rewires enterprise AI agents before they spawn more silos

Microsoft is trying to solve a problem that is already haunting enterprise AI: every new agent starts dumb. No memory of how the business works, no map of where data lives, no built-in sense of what rules apply. At Build 2026, the company responded with two tightly linked pieces, Microsoft IQ and Rayfin, both designed to keep agents from multiplying chaos faster than enterprises can govern it.

The timing matters because the market is clearly moving beyond the old question of whether retrieval-augmented generation, or RAG, is enough. According to VentureBeat's VB Pulse's Q1 2026 RAG Infrastructure Market Tracker, hybrid retrieval intent among organizations with more than 100 employees tripled from 10.3% in January to 33.3% in March. In plain English, enterprises are no longer just trying to expand RAG coverage. They are starting to care about the architecture underneath it, which is where Microsoft is making its play.

Microsoft IQ is the company’s answer on the context side. It expands Fabric IQ, Microsoft’s existing business data context layer, into a broader unified system that combines four sources of context so a developer can connect a new agent in one integration step. Those sources are Work IQ, Foundry IQ, Fabric IQ, and Web IQ. Work IQ captures how the organization operates day to day through email, documents, meetings, and schedules, giving agents a picture of people, teams, and workflows. Foundry IQ manages institutional knowledge by curating and indexing knowledge bases so agents understand what it means to work inside the organization, what rules apply, and what procedures to follow.

Fabric IQ models the live operational state of the business through data, defining entities, relationships, and business rules grounded in real-time signals from Fabric Real-Time Intelligence. Microsoft said the ontologies, the layer that captures that operational context, are expected to reach GA in the coming months. Web IQ adds real-time global context from the web, so agents can see what is happening outside the company alongside internal data. Amir Netz, CTO of Microsoft Fabric, framed that ambition in film terms, saying the green screen of cascading code in "The Matrix" was not atmosphere, but the layer that built the world Agent Smith operated in. "Our job in the world of data is creating reality for agents based on data," Netz told VentureBeat.

That framing gets at the real issue for enterprise buyers. Retrieval can fetch facts, but it does not automatically give an agent the shared business context needed to act correctly inside an organization. Microsoft is betting that if agents are going to behave like useful workers, they need something closer to a common operating picture than a pile of documents and embeddings. Netz put it bluntly: "The agents are going to become highly informed virtual employees," and "That's where the world is heading."

Rayfin tackles the other half of the mess. As agentic coding tools spin up applications faster than anyone can govern them, each one risks becoming another silo outside the data layer entirely. Rayfin is a new open-source SDK and CLI that deploys agent-built applications directly to Fabric as a governed production backend, routing application data into the same platform instead of letting every app wander off and create its own isolated stack. In Microsoft’s setup, application data lands in Microsoft OneLake by default and feeds back into the Microsoft IQ context layer rather than accumulating outside it.

That positioning puts Rayfin up against the Postgres-compatible backends that agentic coding tools typically default to, specifically Supabase and Neon. The differentiator Microsoft is pushing is governance. Rayfin routes the application fleet through Fabric’s unified data and compliance layer, rather than allowing a separate backend sprawl to emerge around each agent-built app. Netz described the relationship as bidirectional: the agent building a Rayfin application draws from the organization’s ontology, and the data that application generates then enriches that ontology for the next agent. For enterprises, that is the promise of compounding context instead of compounding fragmentation.

Microsoft is not alone in chasing this prize. Snowflake announced its own context capabilities this week with semantic capabilities. Pinecone has its Nexus platform, which expands the vector database into a knowledge engine, and Redis has built its Iris context and memory platform. The common theme is clear enough to see from orbit: RAG and model availability are no longer the only bottlenecks. The harder problem is making AI systems aware of the enterprise they are operating in, and then keeping the applications they produce inside a governed perimeter.

That is why Robert Kramer, managing partner at KramerERP, said the enterprise AI challenge is no longer just about model availability. "The real question is whether Microsoft simplifies execution and strengthens trust or adds another layer to an already complex environment," he told VentureBeat. That is the needle Microsoft has to thread. If it works, CIOs and platform teams get a cleaner way to connect agents to business context and deploy agent-built apps without spawning more shadow infrastructure. If it does not, the platform risk is familiar: one more layer, one more abstraction, and one more place for enterprise complexity to hide.