Snowflake, Salesforce, dbt Labs and others push semantic interoperability forward — and with it, the future of metadata-driven, model-first data engineering.
Summary
The Open Semantic Interchange (OSI) Initiative — announced by Snowflake, Salesforce, dbt Labs and partners — introduces an open, vendor-neutral way to exchange business meaning and semantics across data platforms.
For practitioners of Model-Driven Data Engineering (MDDE), this is a long-awaited milestone: it enables model-first, metadata-driven, business-friendly architectures to scale across diverse tools, support AI readiness, and finally move beyond isolated metadata silos.
Why OSI Matters
In our earlier writings — from From Coding to Modeling: The Rise of Model-Driven Data Engineering to Metadata-Driven Data Architecture: Leveraging draw.io and VSCode for Scalable Design and How Universal Data Models Support Business-Oriented Data Modeling — we’ve argued that data engineering must move:
From hand-crafted pipelines to model-first, metadata-driven, business-friendly automation.
But even as tools such as dbt, OpenMetadata and metadata-centric modeling matured, one fundamental challenge persisted:
Every platform speaks its own semantic dialect.
Meaning gets lost in translation. Interoperability remains fragile.
OSI aims to solve this by defining a shared semantic layer — making business meaning portable and consistent across the modern data stack.
How OSI Connects to Our Methodology
1. Model-First Engineering
We design source models and business-friendly mappings first, then generate pipelines. OSI could make these models portable: a logical integration layer or business-shaped model could be exported once and consumed by dbt or Snowflake Cortex without semantic loss.
2. Metadata-Driven Automation
We’ve long said that metadata is not documentation — it’s the backbone of automation (see Metadata as Data — Controlling and Governing Data Models).
OSI provides the standard metadata language needed for governance, lineage, and automation to flow between tools seamlessly.
3. Business-Friendly Mapping
Our drive to make mappings clear to business users — as in Cooking with Data: How Business-Friendly Mapping Brings Clarity to ETL — could now persist end-to-end. OSI makes it possible for conceptual mappings to stay intact through execution.
4. Universal & Business-Oriented Models
We’ve promoted universal data models for years (see How Universal Data Models Support Business-Oriented Data Modeling). OSI makes them shareable industry assets: imagine importing a FIBO or GS1 model in OSI format and using it across design, governance, and transformation.
5. AI-Ready Data
AI thrives on semantic clarity. By providing a shared language for meaning, OSI can feed AI models with trusted, well-described data, reinforcing the principles discussed in Data Contracts: Anchoring Reliability in Model-Driven Data Engineering.
Why This Feels Different
Semantic standards have come and gone.
What’s different now:
- Snowflake brings the cloud-scale data & AI platform.
- Salesforce contributes enterprise and customer-centric semantics.
- dbt Labs represents the transformation layer embraced by engineers and analysts.
Together, these players can move the market — aligning with the metadata-driven, open ecosystem trend our community has been building for years.
Key Takeaways
- OSI could become the universal “semantic glue” connecting modeling, transformation, governance, and AI.
- Model-first, metadata-driven architectures will gain true portability, eliminating brittle hand-offs between tools.
- Business-friendly mappings and universal models will travel further, finally realizing the vision of reusable, business-aligned semantics.
- AI-readiness will no longer be an afterthought — semantic clarity will be built into the stack by default.
- Practitioners should prepare now: document business meaning rigorously, modernize metadata repositories, and look for OSI support in tools.
Further Reading
- From Coding to Modeling: The Rise of Model-Driven Data Engineering
- Metadata as Data — Controlling and Governing Data Models
- Metadata-Driven Data Architecture: Leveraging draw.io and VSCode for Scalable Design
- Cooking with Data: How Business-Friendly Mapping Brings Clarity to ETL
- How Universal Data Models Support Business-Oriented Data Modeling
- Data Contracts: Anchoring Reliability in Model-Driven Data Engineering
A Vision Coming to Life
When I started advocating model-driven data engineering, I was inspired by pioneers such as Hans Hultgren (Ensemble Modeling), Rick van der Lans (data virtualization and metadata-driven thinking), Simon Späti, Roelant Vos, and others who shaped my view of metadata-first architecture.
Their influence — combined with the industry’s evolution from Kimball and Inmon to metadata-driven, business-friendly design — laid the foundation we work with today.
The Open Semantic Interchange Initiative is the clearest signal yet that this vision is becoming industry standard.
It’s not just about AI.
It’s about making data engineering scalable, governed, and business-aligned — exactly what Model-Driven Data Engineering set out to achieve.