Enterprise AI agents continue to create data silos. Microsoft’s answer Build Microsoft IQ and Rayfin.
Every new AI agent your team uses starts from scratch: no memory of how the business works, where the data lives, or any rules that apply. And since agent coding tools crawl applications faster than anyone can control them, each one runs the risk of becoming another silo outside of your data layer entirely. Microsoft is addressing both issues head-on in Build 2026.
According to VentureBeat’s VB Pulse’s Q1 2026 RAG Infrastructure Market Tracker, mixed return intent among 100-plus labor organizations tripled from 10.3% in January to 33.3% in March, a sign that businesses have moved past RAG expansion and are now focusing on infrastructure. Shared business content part retrieval does not resolve.
On the context side, Microsoft is expanding Fabric IQ, its core layer of existing business data, into a comprehensive integrated system called Microsoft IQ, adding three more context sources that include how the organization works, what it knows and real-time real-time landmarks from the web, so that any agent can tap all four as a single foundation. On the application side, Rayfin, a new open source SDK and CLI, runs applications built for agents directly on Fabric as a controlled production backend, moving application data to a common location rather than spinning up new silos.
Amir Netz, CTO of Microsoft Fabric, reached for a movie analogy to explain where the data platform fits. Cascade code green screen inside "The Matrix" it wasn’t space, it was the layer that made up the world in which Agent Smith worked.
"Our work in the world of data creates reality for agents based on data," Netz told VentureBeat.
Microsoft IQ combines four content sources into a single agent base
Microsoft IQ integrates four content sources that until now existed separately, designed so that a developer can connect a new agent to all four in one integration step.
IQ function. It captures how the organization works day-to-day, drawing on email, documents, meetings and schedules to give agents insight into people, teams and workflows.
Foundry IQ. Manages institutional information, configuration and information bases so that agents understand what it means to work within the organization, what rules apply and what procedures must be followed.
IQ fabric. It models the live operational state of the business with data, defining entities, relationships and business rules based on real-time signals from Fabric Real-Time Intelligence. Ontologies, the layer that captures that operational context, are expected to reach GA in the coming months.
Web IQ. It adds real-time global context from the web, giving agents a current picture of the world outside the organization alongside its internal data.
"Agents will be highly skilled workers," Netz said. "This is where the world is headed."
Rayfin routes the applications created by the agent into the same database
Creating shared content solves one part of the problem. Another thing that happens when agents start generating applications. Every new application needs a backend, and without a dominant implementation method each one creates new data out of the context layer entirely.
Rayfin provides an enterprise-grade back end and deploys applications built for agents directly to the Fabric, so application data resides in Microsoft OneLake automatically and flows back into the Microsoft IQ core layer rather than accumulating outside of it.
Microsoft pits Rayfin against Supabase and Neon, Postgres-compatible backends that agent coding tools default to. The difference is management: Rayfin manages the entire application fleet with a unified Data Fabric and compliance layer instead of building individual silos.
Netz described the relationship as two-way. The agent that builds the Rayfin application takes in the organization ontology. The data generated by the application then enriches that ontology for the next agent.
Every big data platform is chasing the same answer, but execution is not guaranteed
Microsoft is not the only platform building a shared context layer for agents. Snowflake announced its context capabilities this week with semantic capabilities. Pinecone has its own Nexus platform that extends the vector database to become a database engine and Redis has developed its own Iris core and in-memory platform.
Microsoft’s approach also reinforces the trend that the RAG and availability model is no longer an issue.
"Fabric IQ and Rayfin are important because the AI challenge for the enterprise is no longer just an availability model," Robert Kramer, managing partner at KramerERP told VentureBeat. "The real question is whether Microsoft is simplifying and strengthening trust or adding another layer to an already complex environment."
