Microsoft is tackling one of artificial intelligence's most overlooked problems: messy, fragmented data that prevents AI agents from working effectively. Under Satya Nadella's leadership, the company is rolling out a suite of tools designed to unify how enterprises manage, label, and share data across their organizations. The goal is to make it possible for AI systems to understand not just raw information, but the business context and rules that govern how organizations actually operate .

Why Is Data Fragmentation Killing Enterprise AI Projects?

At the heart of Microsoft's new strategy is a simple but powerful insight: artificial intelligence cannot perform well when data is scattered, poorly labeled, and disconnected from business logic. Amir Netz, Chief Technology Officer of Microsoft Fabric, explained the scale of the problem in stark terms .

"If I took the smartest person in the world, IQ 180, just graduated from MIT, and I brought them in front of the data lake with 100,000 tables, and I gave them a business question, they will struggle," stated Amir Netz, CTO of Microsoft Fabric at Microsoft.

Amir Netz, CTO of Microsoft Fabric at Microsoft

This challenge becomes exponentially worse for AI agents. While a human expert might eventually navigate a confusing data landscape, an artificial intelligence system lacks the institutional knowledge to understand which tables matter, how they relate to each other, or what business rules govern their use. Microsoft's response has been to consolidate approximately 20 distinct data services into a single platform called Fabric, which now serves 90% of Fortune 500 companies .

The fragmentation problem extends beyond technical architecture. Data is often named and labeled poorly across organizations, making it nearly impossible for AI systems to extract meaningful insights. Microsoft's solution involves creating semantic models that act as a "reference manual" for dense data, helping AI systems understand not just what information exists, but what it means in the context of a specific business .

How Is Microsoft Building AI That Understands Business Rules?

Microsoft's latest announcements reveal a multi-layered approach to solving the data fragmentation problem. The company unveiled two major expansions to its Fabric platform at FabCon and SQLCon 2026 in Atlanta .

• Database Hub: A unified database management system with built-in AI agents and Copilot integration that allows natural language exploration of enterprise data across Azure SQL, Azure Cosmos DB, Azure Database for PostgreSQL, and other sources without requiring users to write code.
• Fabric IQ Enhancements: New semantic and ontological controls that help AI systems understand organizational processes, business rules, and relationships between data entities through a model context protocol (MCP) server.
• Copilot Cowork: An agentic AI feature that can autonomously plan and execute tasks across Microsoft 365 applications like Excel, Outlook, Teams, and PowerPoint, with users simply describing what they need done.

The concept of "ontology" is central to Microsoft's strategy. Unlike traditional data management, which focuses on where information is stored, ontology describes how a business actually operates. Netz explained this using the example of an airline .

"The ontology is describing your business in the most operational aspect of the business. How are things related? What are the rules? What are the policies? What are the actions you can take? What can I do with a plane? I can assign a crew to a plane. I can divert the plane. I can ground a plane. I can load the plane with luggage," explained Amir Netz, CTO of Microsoft Fabric.

Amir Netz, CTO of Microsoft Fabric at Microsoft

This approach is essential because AI agents need to understand not just what data exists, but what they are allowed to do with it. Microsoft expects 1.3 billion AI agents to go live by 2028, and without clear organizational guardrails, these systems could make decisions that violate business policies or ethical constraints .

What New Research and Comparison Tools Is Microsoft Releasing?

Beyond data consolidation, Microsoft is advancing how AI systems themselves work together to produce better results. The company announced Wave 3 of Microsoft 365 Copilot on March 30, 2026, which includes several innovations designed to improve research and analysis .

A new "Critique" feature uses multiple AI models from companies like Anthropic and OpenAI to improve research quality. The system separates the generation phase from the evaluation phase, with one model creating a draft and a second model reviewing and refining it. In the DRACO benchmark, which measures accuracy, completeness, and objectivity in research, this approach achieved a 13.8% higher score compared to single-model approaches .

Microsoft also released "Model Council," a tool that runs multiple AI models simultaneously on the same research question. Each model generates a full report, and then a third "judge" model provides a detailed summary comparing where the models agree, disagree, and offer unique insights. This allows researchers to see how different AI systems approach the same problem and make more informed decisions about which insights to trust .

How Can Organizations Prepare for Agentic AI Deployment?

As Microsoft rolls out these tools, organizations need to think carefully about how they structure their data and define their business rules. The transition from traditional software to agentic AI requires a fundamentally different approach to data governance .

• Map Your Data Relationships: Organizations should create semantic models that describe how their data connects and what it represents in business terms, not just technical terms.
• Define Organizational Ontology: Clearly document business rules, policies, and decision-making frameworks that AI agents need to understand, such as how to prioritize competing business goals like profitability versus safety.
• Establish AI Agent Governance: Set up oversight mechanisms and behavioral constraints using open standards like MCP and agent-to-agent (A2A) protocols to ensure AI systems operate within acceptable boundaries.
• Consolidate Data Sources: Use tools like Fabric's OneLake and Mirroring to create a unified view of data across systems, eliminating the need to duplicate information while maintaining low-latency access.

Netz stressed that organizations will need human managers overseeing AI agents, with lower-level database administration tasks increasingly automated. However, the critical decisions about business priorities and ethical constraints cannot be delegated to machines .

"Think back to British Airways: they want to be profitable, but they also want to have customer satisfaction and they also want to take off and land the plane safely. Which one takes precedence over which? How much are you willing to sacrifice in safety to get more profitability? These are things that MCP tools don't tell you, because these are the things that we expect employees to understand," noted Amir Netz, CTO of Microsoft Fabric.

Amir Netz, CTO of Microsoft Fabric at Microsoft

Microsoft's strategy reflects a broader shift in how enterprises approach artificial intelligence. Rather than deploying AI as a standalone tool, companies are learning that AI works best when it is deeply integrated with organizational data, processes, and values. By consolidating data management and making it easier for AI systems to understand business context, Microsoft is addressing what may be the most significant bottleneck preventing AI from delivering real value at scale .