When we launched AtScale over a decade ago, the idea of a universal semantic layer was ahead of its time. Today, it’s no longer a vision, and it’s the foundation for modern analytics and enterprise AI.
At the 2025 Semantic Layer Summit, I had the honor of kicking off the keynote to an audience of data leaders, architects, and innovators who are rallying around a shared belief: AI and analytics should be open, interoperable, and grounded in trust. This year’s Summit wasn’t just about new product features (though we launched plenty). It was about something bigger: the rise of open semantics and how open models, formats, and collaboration will shape the next era of business intelligence.
Let’s talk about why.
AI Without Context Is Reckless
As I shared in my keynote, generative AI has reached a tipping point. It’s in our dashboards, embedded in workflows, and increasingly shaping business decisions. But as many organizations have learned the hard way, GenAI without guardrails is dangerous.
In our own testing with TPC-DS, we found that large language models (LLMs) are incorrect over 80% of the time when left to their own devices. But when grounded in a semantic layer that encodes business logic, joins, metrics, and relationships, we achieved near 100% accuracy across tools like Databricks Genie and Snowflake Cortex Analyst.
The takeaway is clear: AI is only as useful as the context it understands. And that context lives in the semantic layer.
The Problem with Proprietary Semantics
We’re now seeing an explosion of proprietary semantic layer initiatives, including Snowflake, Google Looker, Microsoft, Databricks, and Tableau, all building closed ecosystems. While that signals the category’s maturity, it also reintroduces the problems of vendor lock-in, data silos, and tool fragmentation.
That’s why we open-sourced SML (Semantic Modeling Language), the industry’s first open, implementation-neutral modeling standard. SML allows teams to define metrics, relationships, and business logic once and use them everywhere.
Because if we want semantic consistency to scale, we need a common language.
Why Open Semantics Win
Dael Williamson, Field CTO at Databricks, delivered a powerful presentation at the Summit that echoed this urgency. He explained it like this:
“As organizations mature their AI strategies, they’re realizing the most valuable data lives at the top of the stack, closest to the human layer, and it’s completely locked in. That’s where the semantic layer lives. That’s where the meaning, the context, and the competitive differentiation are. But it’s also where there’s the most lock-in.”
Dael went on to explain that enterprise AI isn’t just about big models. It’s about domain-specific intelligence, small, focused, context-rich systems that understand your business. And that requires structured semantics. The answer, he said, is open source:
“The best technology for solving complex problems is open source. We’ve seen it at the operating system layer, in data formats, and now in semantics. It’s time to open the layer closest to decision-making.”
AtScale’s open approach isn’t just about sharing code. It’s about creating a community and a standard for how business logic gets defined, governed, and reused across tools and teams.
New Innovations: Building Toward an Open Future
To accelerate this vision, we announced several new product capabilities at the Summit that are grounded in openness and composability:
MCP Server: Semantic Context for AI Agents
To fully realize the promise of AI agents acting on business data, we introduced the AtScale MCP (Model Context Protocol) Server—a new interface purpose-built to serve governed semantic logic directly to AI agents and LLMs.
The MCP Server allows autonomous systems to query AtScale models using natural language or structured prompts and receive responses grounded in business-defined metrics, hierarchies, and filters. This ensures that AI agents reason and act with the same trusted definitions used by analysts and executives, removing ambiguity and improving explainability.
One-Click Modeling
With one click, users can now automatically generate complete semantic models, dimensions, metrics, joins, and all, using AI. This removes barriers for non-technical users and makes standardized modeling accessible to everyone.
Composite Modeling
This feature allows teams to reuse and link semantic models across domains. You can combine your revenue model from sales and your cost model from finance to build new, governed cross-domain KPIs like profitability without duplicating work.
PGWire (Postgres) Support
We now support the inbound PostgreSQL wire protocol, meaning AtScale can connect to virtually any BI tool, including ThoughtSpot, Superset, Sigma, and more, without custom integrations. This makes AtScale’s semantic models truly interoperable across your data stack.
Native Power BI Tabular Support
We’ve deepened our Power BI integration to support native tabular models and DAX, enabling users to work with familiar expressions while inheriting governance, definitions, and real-time updates from the semantic layer, no more imports or stale dashboards.
These aren’t just features. They’re building blocks of a composable, open, and governed analytics ecosystem that lets data teams and business users move faster, together.
The Road Ahead: Interoperability Is the Standard
In his session, Dael described a future where organizations can collaborate seamlessly across domains, clouds, and tools, sharing semantically consistent definitions of things like “revenue” or “customer” across teams. But to do that, the context layer, our semantic layer, must be open and interoperable.
That’s exactly where we’re heading.
We’ve already made it possible to deploy AtScale semantic models across Databricks, Snowflake, and other platforms, so no rebuilding is required. And we’re just getting started. With open SML, cross-platform support, and rich metadata APIs, the goal is to unify semantics across the enterprise, not lock them away in one vendor’s ecosystem.
Why This Matters Now
In the end, this is about unlocking the value of your data—not just for data scientists or analysts, but for everyone.
When we talk about “open semantics,” we’re talking about removing barriers to insight. We’re talking about giving every person in your organization, from the CFO to a store manager to a digital product owner, the ability to ask questions, get answers, and make decisions confidently.
We’re talking about shifting from BI as a bottleneck… to BI as a network of intelligence.
The future isn’t just AI-powered, it’s context-powered, collaboration-powered, and open-powered. Let’s build it together.
Want to join the open semantics movement?
Explore our SML GitHub repo, try out One-Click Modeling in your own environment, or watch the full product keynote on demand from the 2025 Semantic Layer Summit.
The future is open. And it starts now.
SHARE
Case Study: Vodafone Portugal Modernizes Data Analytics