SAP governance is becoming an AI control plane, not a compliance afterthought
SAP governance in the AI era is the set of technical and policy controls that manage how AI agents, analytics platforms, and integration tools access, interpret, and act on SAP and non-SAP data across the SAP Business Data Cloud and related systems, with the goal of protecting compliance, trust, and operational safety while still enabling new automation and insight use cases.
What stands out this month is how three vendors are quietly redrawing that control plane. Microsoft is turning Power Platform into an action-aware SAP connector governance layer for AI agents, CData is pushing governed connectivity and AI developer tools into SAP Business Data Cloud, and Collibra is stitching governance directly into Snowflake and Databricks, the engines under BDC. These moves are not minor feature updates; they are evidence of a clear bet: the next phase of enterprise data governance will live inside the tools that run AI, not in a separate, ornamental framework. Enterprise architects who treat this as “just” integration will end up with agents they cannot explain, audit, or contain.
Microsoft turns SAP connector governance into an action-level AI safety rail
Microsoft’s June Power Platform releases are a direct answer to the new compliance surface created by AI agents reaching into SAP systems. On June 4, 2026, the company announced general availability of Advanced Connector Policies for Power Platform, followed a week later by a broad update to SAP guidance in the June 11 feature release. Together they reposition Power Platform as a governed integration layer for SAP environments running AI agents, not just a low-code workflow tool.
Advanced Connector Policies replace the classic data loss prevention model with allowlist-based governance at the action and Model Context Protocol (MCP) server level. Administrators can now decide not only which connectors are allowed, but which SAP connector actions and MCP servers AI tools are permitted to use. That matters when both the SAP ERP connector and SAP OData connector now sit under the same action-level framework. The practical impact is blunt: AI builders lose some freedom, but Power Platform admins gain precise control over what agentic access to SAP is acceptable, turning SAP connector governance into a meaningful safety rail instead of a checkbox.
CData’s AI developer tools make SAP Business Data Cloud about real connectivity, not diagrams
If Microsoft is tightening the front door into SAP, CData is widening the governed corridor that connects SAP Business Data Cloud to everything else. CData has been embedded in SAP BDC since 2025, giving the platform governed, real-time access to hundreds of non-SAP sources so BDC can unify governance, analytics, and AI workloads on a single platform. That foundation matters because SAP’s own research shows only 3% of organizations have achieved a unified, governed data layer, while 38% remain in siloed or ad hoc integration states.
On June 23, 2026, CData shipped a set of AI developer tools and earlier in the year expanded the leadership team that runs its data layer. The release centers on MCP again: a free Connect AI Developer Edition to give AI teams governed data access without a full integration stack, an open-source Connect AI Python SDK for agentic workflows, and a CData CLI that lets developers query enterprise data sources from the command line. The stated goal is to replace the ungoverned, brittle pipelines that slow most AI projects. In other words, AI developer tools for SAP are becoming governance tools by design, not afterthoughts.
Collibra binds governance into Snowflake and Databricks at the heart of SAP BDC
While Microsoft and CData focus on connectors and AI tooling, Collibra is moving where many governance vendors talk but rarely reach: inside the engines that power SAP Business Data Cloud. SAP BDC, the fully managed data layer SAP and Databricks introduced in 2025, runs on SAP Databricks and connects to Snowflake through zero-copy data sharing. Collibra is already SAP’s chosen governance partner for that platform, and its June wins strengthen that position.
On June 2, Collibra and Snowflake expanded their partnership to enable bidirectional exchange of governed business context and semantics between the two platforms. That integration will synchronize governed context and semantic definitions, grounding Snowflake Cortex Analyst and Cortex Agents in trusted metadata, with broader availability slated for Q3 2026. Two weeks later, on June 16, Databricks named Collibra its Governance Partner of the Year, as Collibra’s bidirectional integration with Unity Catalog, AI Command Center, and a new MCP server went live on the Databricks Marketplace. When only 12% of organizations have reached the level of automated governance required to support AI-driven workloads, embedding Snowflake Databricks governance this deeply is less award fodder and more a statement: AI governance will follow the engines, not the slideware.
Architects must treat SAP AI governance as a product decision, not a policy document
Across Microsoft, CData, and Collibra, a pattern is clear: governance is moving from central committees into the products that run AI. Power Platform’s Advanced Connector Policies show that governed AI connectivity to SAP is now a selection criterion; administrators are judging platforms on action-level and MCP server controls, not connector allowlists alone. CData’s MCP-based connectivity layer makes SAP BDC’s claim to unify governance, analytics, and AI workloads contingent on how well enterprises extend their data governance policies to cover MCP-sourced data before prototypes reach production.
Meanwhile, Collibra’s deeper integration with Snowflake and Databricks gives CIOs a more credible answer to the question of whether their governance tools truly cover the engines SAP BDC runs on. Architects should read these moves as a warning. When only about 20% of organizations have fully integrated systems with real-time data flow, while the average SAP shop connects 36 applications using four or more tools, agents built on ungoverned foundations will fail loudly. SAP customers are not buying SAP BDC to admire a clean architecture diagram but to run agents on trusted data, and those agents will expose every gap in compliance, inventory visibility, and AI agent control.
