During FabCon Poland 2025, Microsoft presented a wide range of updates and new capabilities in Microsoft Fabric, including the latest announcements unveiled at Microsoft Ignite in November. The material focused on the evolution of the platform’s architecture, new database services, data integration mechanisms, AI capabilities, and the ongoing development of the security and governance layer. These changes will have a direct impact in the coming quarters on how data platforms built on Fabric are designed and operated.
Microsoft Fabric is positioned as a platform that brings order to both structured and unstructured data, bridges transactional and analytical workloads, and creates the conditions for the secure and responsible use of AI. In this context, Fabric is not merely an analytics tool, but a foundational layer for building systems, applications, and AI agents as organizations evolve over the next quarters.
OLTP and OLAP in Microsoft Fabric: A New Approach to Data ArchitectureOLTP i OLAP w Microsoft Fabric – nowe podejście do architektury danych
One of the most significant changes introduced in 2025 is the convergence of OLTP and OLAP workloads within a single platform. Microsoft Fabric introduces native database services that enable transactional system support without the need to maintain separate infrastructure.
SQL Database in Fabric, now generally available, serves as the transactional counterpart of SQL Server for ERP or CRM systems. It is complemented by Cosmos DB in Fabric, designed for semi-structured data such as e-commerce events or user behavioral data. An announced upcoming capability is HorizonDB, a database engineered for AI and RAG scenarios, including vector storage and semantic query processing.
A key architectural element is data mirroring, which allows replication of data from SQL Server, Azure PostgreSQL, or Oracle without traditional ETL processes. A particularly important conceptual shift is mirroring to SAP Business Data Cloud, where not tables but business objects are replicated together with their existing logic and measures.
OneLake, Lakehouse, and Dataflows Gen2: A Single Data Flow from Transactions to Analytics
At the core of the platform remains OneLake as the shared data storage layer underpinning the Lakehouse model. This is where transactional, analytical, and streaming data are brought together, eliminating the need to build separate silos.
Dataflows Gen2 have been extended with AI Transformations, enabling data transformations using natural language. Examples include automatic sentiment analysis or generating personalized content based on selected columns, without relying on external AI services. The pipeline layer is complemented by dbt jobs, allowing transformations to be implemented consistently without leaving the Fabric platform.
At the operational layer, Copilot Right Approval has been introduced, ensuring that administrative actions recommended by Copilot are only executed after explicit user approval.
Fabric IQ, Data Agents, and Operations Agents: Microsoft Fabric as an Agent Platform
One of the most important development directions for Fabric is its support for AI agents, both interactive and background-running. A central role is played by Fabric IQ, a data ontology layer that describes an organization’s business model in a way that is understandable to agents.
This ontology is built on entities, properties, rules, actions, semantics, and business goals. As a result, agents operate not on raw tables, but on business concepts such as sales, margin, or customer. On this foundation run capabilities like Power BI Chat with Your Data, Fabric Data Agents supporting data teams, and Fabric Operations Agents, which automatically respond to events and communicate with users, for example through Microsoft Teams.
Additional components include User Data Functions, APIs, and GraphQL, which allow application-level actions to be triggered directly from data stored in Fabric.
Security and Governance in Microsoft Fabric: OneLake Security and Microsoft Purview
In 2025, Microsoft completed its Unified Data Platform vision by introducing OneLake Security as the central security management layer. Row-level, column-level, and object-level security policies are defined in a single place and inherited across all engines, including SQL, Spark, and Power BI.
This is complemented by Microsoft Purview, which provides data labeling, DLP mechanisms, and sensitivity controls. Purview Unified Catalog also enables the maintenance of business descriptions for data and models, supporting both governance processes and the operation of AI agents.
What These Changes in Microsoft Fabric Mean for Data Leaders and IT Decision Makers
The updates presented at FabCon Poland 2025 clearly show that Microsoft Fabric is evolving toward an operational and agent-centric platform, rather than remaining purely an analytics tool. Combining OLTP, OLAP, AI, and governance in a single environment reduces architectural complexity and shortens the path from data to decisions.
