By the end of 2025, the conversation around analytics stopped being about dashboards and started sounding a lot more like operations. The rise of autonomous and semi-autonomous agents put a sharper edge on an old truth: AI only becomes an enterprise capability when the underlying data is trusted, discoverable, and defensible.
Microsoft Fabric and Microsoft Purview spent 2025 building toward that reality from opposite (but increasingly overlapping) sides of the house. Fabric pushed the platform forward—unifying workloads, expanding OneLake, and adding new intelligence and database capabilities designed for AI-era workloads. Purview tightened the governance and security loop—making data quality, cataloging, risk visibility, and policy enforcement feel less like a separate initiative and more like part of the daily flow.
This year-in-review walks through the story they told together: how Microsoft Fabric expanded the data platform into a more complete data estate, how Microsoft Purview reframed governance for the AI and agent era, and where the two converged into a more practical, enterprise-ready operating model.
The headline story they told: unified data is necessary, governed data is usable
Fabric’s 2025 arc was about closing gaps—between operational and analytical data, batch and real time, structured and unstructured, and human insight and machine action. One visible marker of that traction: Microsoft publicly positioned Fabric as a rapidly adopted platform, reporting more than 28,000 customers (including a large share of the Fortune 500) using it by late 2025.
Purview’s 2025 arc was about making that growth safe. It leaned into a federated governance posture—central standards, distributed ownership—because modern data estates (and AI usage patterns) are too wide for a single team to curate manually. That federated framing is baked into the Unified Catalog story: a single SaaS experience meant to reduce tool switching, organize data by business-aligned governance domains, and connect governance to value creation rather than only risk avoidance.
A simple way to summarize the year is this:
- Fabric made it easier to create and activate data products and AI-ready assets inside one platform.
- Purview made it easier to find, understand, and control those assets across the estate.
- The two products increasingly met in the middle, especially around OneLake security, Fabric item protections, and AI interaction governance.
- The “why now” was unmistakable: enterprise AI demanded data governance that can keep pace with agentic AI.
Fabric’s 2025: expanding from a platform into a data estate for AI
The database layer arrived as a first-class Fabric citizen
One of the most consequential shifts in 2025 was Fabric becoming more explicitly “developer-first,” not just “analytics-first.”
At Ignite 2025, Microsoft positioned Fabric databases as generally available—bringing SQL database and Cosmos DB together in a unified SaaS experience, with enterprise security and native AI integration patterns such as support for vector data and RAG-style applications.
On the product side, Fabric also called out the SQL database in Fabric as generally available (built on the Azure SQL Database engine), and Cosmos DB in Fabric as generally available, including additions such as vector indexing and search.
From a year-in-review lens, this mattered because it shrank the distance between:
- operational data creation,
- analytical modeling and reporting, and
- AI applications that need both “system of record” and “system of insight” in the same governed boundary.
OneLake became more interoperable—and more “everyday”
Fabric’s foundational bet—OneLake as the backbone—kept broadening in practical directions.
On the interoperability side, Microsoft described expanded OneLake connectivity across platforms such as SAP, Salesforce, Azure Databricks, and Snowflake, and framed mirroring as a way to reduce ETL burden and make operational data “instantly ready” for analytics and AI.
On the “everyday” side, December updates highlighted OneDrive and SharePoint shortcuts in OneLake—a signal that the boundary of “data” was widening to include the documents and folders that actually power business work.
This is the kind of change that sounds small until it isn’t: bringing Microsoft 365 content closer to the governed analytics plane is one of the fastest ways to increase adoption and increase risk—meaning it only works well when governance is welded to the workflow.
The intelligence layer took shape: Fabric IQ, operations agents, and data agents
At Ignite, Microsoft introduced Fabric IQ as a semantic foundation, explicitly positioned as a force multiplier rather than a replacement for existing data estates.
Microsoft’s broader Fabric narrative added important detail: Fabric IQ is a workload designed to map datasets to real-world entities and relationships, built on semantic model technology, and paired with operations agentsthat can monitor real-time sources and take proactive action.
At the same time, Fabric leaned into data agents as an enterprise bridge between data and action. In mid-year, Fabric highlighted a preview integration between Fabric data agents and Microsoft Copilot Studio.
Microsoft also described data agents becoming embeddable in Microsoft 365 Copilot and even acting as hosted Model Context Protocol (MCP) servers—an architectural hint that Fabric expects to be a grounding and reasoning layer for many AI experiences, not just Fabric-native ones.
Purview’s 2025: governance and security moved closer to where work happens
Unified Catalog: governance organized like a business, not like a storage account
Purview’s Unified Catalog story is easy to misread as “a nicer catalog UI.” In 2025 it was more than that.
The Unified Catalog experience was described as rolling out regionally and built around governance domains, data products, access policies, critical data elements, and glossary terms—paired with AI-powered search and workflows designed to make discovery and access more self-service.
It also tied governance directly to business outcomes through mechanisms like OKR alignment and data health/quality scoring across assets, products, and domains.
This is where data governance matured in tone: it wasn’t only “control,” it was “coordination.”
DSPM evolved for the AI era: posture, workflows, and agent awareness
Purview’s Data Security Posture Management (DSPM) narrative sharpened through 2025. The Microsoft Learn release notes described a new version of DSPM rolling out in preview—bringing earlier DSPM experiences together and adding outcome-based guided workflows, posture reports, and remediation framing.
At Ignite 2025, Microsoft also framed Purview as expanding its data security and compliance capabilities for agents, including new detections and risk analysis for agent actions (in preview).
The subtext here was clear: “AI governance” is no longer a side policy. It is becoming a core part of data security posture.
Compliance workflows got cleaner: unified eDiscovery and broader retention coverage
Purview’s “plumbing” changes mattered too, especially for organizations that live in audits, investigations, and retention policies.
In May 2025, Microsoft introduced changes that transition Content Search and classic eDiscovery (Standard) into a unified Purview eDiscovery experience.
Meanwhile, Purview expanded retention policy locations to explicitly include Copilot experiences (including Copilot in Fabric) and other AI applications—an important signal that Microsoft expects AI interaction data to be treated as governed records, not ephemeral chat.
Where Fabric and Purview converged: governed activation, not governance theater
If 2025 had a “joint release note,” it might have been this: governance needs to show up inside the same workflow where data gets built and used.:
- Information Protection policies for Fabric items and DLP for structured data in OneLake were described as generally available, aimed at preventing oversharing and enforcing protection policies where Fabric work actually happens.
- Insider Risk Management indicators for Power BI were described as generally available for Fabric, extending detection signals into BI behaviors such as viewing, downloading, exporting, and label activity.
- Data Risk Assessments for Fabric were positioned as supporting discovery of overshared Fabric data (preview), focusing on the top accessed workspaces.
- Data security and compliance controls for Copilot in Power BI were described as generally available, including the ability to surface prompt/response risks in DSPM for AI reporting and govern interactions through audit, eDiscovery, and retention.
Later in the year, the governance story expanded beyond Fabric itself and into the Microsoft 365 admin surface. Microsoft described Purview value being integrated directly into the Microsoft 365 admin center (public preview) and described extensions to DLP for Copilot to safeguard prompts (public preview), including real-time controls to prevent responses when prompts contain sensitive data.
The practical takeaway from this convergence is not that “governance got better.” It’s that governance got closer.
What this year suggests for 2026: the winning pattern is “platform + policy” as one motion
Fabric and Purview did not spend 2025 building two separate narratives. They spent it tightening a loop:
- Fabric creates a unified place to engineer, analyze, and operationalize data and AI.
- Purview provides the mechanisms to discover, control access, monitor quality, and reduce security risk—especially as AI agents amplify both opportunity and exposure.
Organizations that treat these as separate tracks tend to hit familiar failure modes: fast prototypes that can’t be trusted, or rigorous governance programs that can’t scale. The 2025 pattern is pointing toward a different operating model—one where governed activation is the default.
The action moving into 2026 is not “add more tools.” It’s to tighten the integration points already available: make OneLake the common substrate, make catalog and data products the shared language, and make security posture and DLP enforcement visible in the same places teams already work.
