Semantic Layer Governance vs. Portability: Why Moving Your Metrics Isn't the Same as Governing Them

Quick Answer 

• Semantic layer portability and governance solve different problems. Portability enables metric definitions to move between tools using standards like OSI, while governance ensures those definitions are consistently enforced with the correct security, business logic, and access controls at query time.

• Enterprise AI requires runtime governance, not just portable semantic models. AI agents can read shared metric definitions, but without centralized enforcement of row-level security, fiscal calendars, currency logic, and audit controls, AI outputs can become inconsistent, inaccurate, or non-compliant.
• Strategy Mosaic provides centralized semantic governance across BI tools and AI agents. By enforcing business logic, security policies, and governance rules at the semantic layer, organizations can deliver trusted, consistent metrics and AI-driven insights across every application, dashboard, and autonomous agent.

When the Open Semantic Interchange launched at Snowflake Summit in June 2026, the announcement carried 54 vendors and a compelling promise, a common YAML format for semantic model definitions that any participating tool can read, write, and exchange. No more custom connectors between every platform pair. No more duplicated metric definitions in each BI tool. One format standard to move business logic across the ecosystem.

Format portability is genuinely useful. But the framing of "open format equals consistency" has a gap that every enterprise analytics team building for AI should understand. Moving a metric definition from one tool to another is not the same as enforcing it once it gets there.

A portable semantic layer can express its metric definitions, dimensions, joins, and relationships in a vendor-neutral schema that participating tools can import and export. OSI's YAML-based specification does exactly this, it provides a structure for describing a semantic model that any supporting tool can read.

Before format standards, moving metric definitions between tools meant rebuilding them by hand. OSI automates the exchange. A metric defined in dbt can now be expressed in an OSI-compatible format and imported by Tableau, Snowflake, Strategy, or any other supporting platform without manual reconstruction.

What portability does not address is what happens after the import. The format specifies how to describe a metric. It does not specify how to govern or execute it.

Semantic layer governance is a runtime enforcement concern. A governed semantic layer ensures that when a user queries a metric through any connected tool, the query executes the correct access controls, correct business logic, and correct data filters applied for that user at that exact moment and context.

Governance covers row-level and column-level security enforced at query time. It covers fiscal calendar logic calculated consistently whether the consumer is a Power BI dashboard, a Tableau sheet, or an AI agent. It covers the audit trail that records who accessed what data, when, and through which tool.

The key distinction: governance is active and enforced, portability is passive and descriptive. A format standard describes a metric, governance ensures that metric behaves correctly everywhere it is used.

If you export your revenue metric definition via OSI and import it into a BI tool that supports the format, does that BI tool automatically apply your row-level security rules?

OSI defines how to describe the metric. It does not define how each consuming tool must enforce access controls on it. A tool that imports an OSI-formatted semantic model can apply its own security model, or none at all. The metric definition travels, the governance policy does not travel with it.

The same applies to fiscal calendars, multi-currency logic, and conditional calculations. A YAML file can describe that a metric uses a fiscal calendar offset. Whether each consuming tool calculates that offset correctly is a function of that tool's implementation, not the format standard.

Format standards advance interoperability at the definition layer. They do not resolve the execution gap.

AI agents raise the governance stakes significantly. When a BI analyst in Power BI queries revenue by region, a misconfigured access control produces a wrong report, which is a potentially costly error. When an AI agent queries the same metric autonomously, processes the result, and surfaces it in a customer-facing recommendation, a governance gap becomes a compliance and liability issue.

AI agents do not pause to verify whether their data access is authorized. They execute against whatever semantic layer they are pointed at. If that semantic layer applies governance at the tool level, inconsistently, or not at all, the agent's output reflects that. Consistent, centrally-enforced governance is not optional for agentic analytics, it is the safety model the entire stack depends on.

Portability standards help AI agents find and read semantic definitions. Governance determines whether those definitions are trustworthy when the agent executes them.

Strategy built Mosaic to address governance as a runtime problem, not a format problem. Mosaic connects to 200+ data sources and enforces a single governance policy across every connected BI tool and AI agent simultaneously. Fiscal calendars, multi-currency logic, row-level and column-level security: these are defined once in the semantic layer and applied automatically at query time, regardless of which tool submits the query.

When an AI agent queries fiscal-quarter revenue by sales territory, Mosaic applies the fiscal calendar offset, the currency conversion for the relevant territory, and the row-level filter for that agent's authorization scope. That logic does not live in the BI tool. It lives in the semantic layer, and it travels with every query.

Mosaic Sentinel, the governance intelligence layer, monitors every data access event in real time. Risk management, audit and compliance, and usage insights modules give data governance teams a complete view of who accessed what, through which tool, and when. That audit trail covers dashboards, reports, human and agentic access alike.

Format portability and runtime governance can coexist. OSI advancing the interchange standard is a positive development for the ecosystem. But governance is not delivered by the format. It is delivered by the semantic layer that enforces it.

The semantic layer ecosystem is maturing, and format standards like OSI reduce real friction in moving metric definitions between tools. But the question enterprises should ask is not "can my semantic models be exported?" It is "are my semantic models enforced consistently everywhere they are consumed?"

Answering that second question requires governance at runtime, not portability at format level.