Data Powers AI, Context Makes it Intelligent: Key Takeaways from FabCon '26
Our team members Diego Prinzi and Marcos Monguillot attended the FabCon ‘26 conference in Atlanta, and they brought exciting news and insights about the pivotal shift in the data landscape.
As the industry mantra suggests, "Data powers AI, but context makes it intelligent". Central to this evolution is OneLake, the unified data lake designed for the era of AI.
OneLake: The Foundation of Unified Data
OneLake serves as the single, logical data lake for the entire organization. It eliminates data silos by providing a unified storage layer where all data—structured, unstructured, real-time, and graph—resides.
- Zero-Copy Integration: OneLake allows for "shortcuts" to data in other clouds like AWS S3 and GCP, or services like Azure Databricks and Snowflake, enabling analytics without moving or duplicating data.
- Direct Lake Mode: This breakthrough storage mode allows Power BI to query massive datasets in OneLake directly without the need for traditional data imports, ensuring performance is fast, real-time, and efficient.
- Unified Security: Security permissions, including Row-Level (RLS) and Column-Level Security (CLS), are respected across the platform, ensuring that AI agents only access authorized data.
AI Functions: Intelligence in a Single Line
A major innovation introduced was AI Functions, which are one-line transformations that apply Large Language Models (LLMs) directly to your rows of data. These functions allow SQL and Python users to perform high-level intelligence tasks without complex API setups or key management.
- Key Capabilities: Built-in operations include ai.summarize, ai.translate, ai.classify, ai.extract, and ai.analyze_sentiment.
- Broad Integration: AI Functions are available across Fabric experiences, including Notebooks, Dataflows Gen2, and the Fabric Data Warehouse.
- Multimodal Support: Beyond plain text, AI Functions now support images and PDFs, allowing users to turn a folder of unstructured files into a structured table in a single call.
The Critical Importance of Context for AI
A recurring theme was that AI agents do not fail for a lack of data; they fail for a lack of meaning. Without business context, AI agents are "powerful but blind," able to process data but unable to reason about it effectively. By providing a unified intelligence layer through Fabric IQ, organizations can ground AI in trusted business knowledge, transitioning from simple retrieval to complex, autonomous reasoning.
Ontology: The Semantic Backbone of the Enterprise
At the heart of Fabric IQ is Ontology, a structured model that formalizes business concepts and relates them directly to your data. It is a live representation of how a business interprets its operations.
- Entity Types: Reusable models of real-world concepts like "Customer" or "Product" that standardize naming across the organization.
- Relationships: Ontology defines how entities interact—for example, a "Customer" places an "Order"—enabling AI agents to perform multi-hop reasoning across complex datasets.
- Rules & Actions: Conditions that define normal vs. abnormal states, allowing Operations Agents to provide 24/7 proactive monitoring and trigger autonomous actions.
Operationalizing Intelligence
The power of a well-defined ontology is its ability to power Data Agents—virtual analysts that handle reactive, ad-hoc Q&A by reasoning over business terms rather than raw database columns. By shifting from reactive dashboards to proactive, autonomous agents grounded in a unified semantic layer, enterprises can achieve true AI-driven transformation.