2026 is shaping up to be the year enterprises finally accept that semantic layers are no longer optional.
As organizations scale AI initiatives and rationalize their analytics stacks, a familiar problem keeps resurfacing. The same business question produces different answers depending on which tool or AI agent you ask. Dashboards disagree. AI assistants return inconsistent metrics. Trust erodes quickly. That is why analysts and practitioners increasingly view governed semantics as foundational infrastructure for both BI and AI.
The industry’s growing focus on open semantics, including initiatives such as the Open Semantic Interchange (OSI), is an important and positive signal. It reflects broad recognition that business logic must travel across tools and systems, rather than remain locked inside individual platforms.
The real challenge now is how to make openness work at true enterprise scale.
The Architectural Reality
Semantic layers are not new. What has changed is where semantic logic lives.
Over time, organizations have embedded business definitions inside BI tools, data platforms, notebooks, and now AI systems. Each environment implements metrics slightly differently. Definitions drift. Governance weakens. Teams spend more time reconciling numbers than using them.
This becomes especially visible as companies move from BI to AI. Ask a dashboard for Q4 revenue, and you get one number. Ask an AI agent the same question, and you get another. Both may query the same data platform, but they are not using the same semantic definitions.
This goes beyond data or tooling. This is an architectural problem.
Why Interoperability Matters and Why Experience Does Too
OSI focuses on semantic interchange, making it easier for semantic definitions to move between tools and representations. That is a meaningful step forward, especially in heterogeneous enterprise environments where no single platform owns the entire analytics or AI stack.
Interoperability reduces friction and acknowledges a reality enterprises already face: semantic logic is fragmented across systems.
What AtScale’s experience adds is perspective on what happens after interchange.
For more than a decade, we have worked with large enterprises such as Home Depot to operationalize semantic layers across BI, analytics, and now AI. That work includes deep expertise in multidimensional analytics and semantic patterns that underpin many enterprise OLAP environments. This experience matters because it directly informs how complex business logic is preserved during modernization.
What we have learned is that the hardest problems are not basic metric definitions. They are rooted in real-world complexity:
- Fiscal calendars that do not follow standard months
- Product and customer hierarchies that evolve over time
- Aggregate-aware calculations that must be correct at any level of detail
- Security and governance rules that must be enforced consistently
Most semantic initiatives struggle with these executional details.
AtScale’s track record includes helping enterprises migrate from legacy proprietary stacks such as SSAS, Cognos, and Oracle. These systems often restrict business users and fail to scale in cloud environments. Companies like Home Depot have moved from outdated OLAP cubes to modern, governed semantic infrastructure that supports everything from Excel to AI agents, without sacrificing correctness or control.
To address these challenges, we developed the Semantic Modeling Language (SML), an open-source, YAML-based language that treats semantic models as code. Models are composable and can be versioned, tested, and governed through standard engineering workflows. This foundation has proven critical as semantics move from dashboards into AI systems, where deterministic behavior and trust are non-negotiable.
AI Raises the Stakes
AI makes semantic consistency more important, not less.
Many enterprise AI initiatives stall because the systems behind them cannot be trusted at scale. AI agents need access to governed, consistent semantic logic without ambiguity or drift.
Without shared approaches to semantics, every AI platform risks re-implementing business logic slightly differently. The result is new silos and new sources of inconsistency. Open, governed semantics are essential to closing the gap between AI pilots and production systems.
Looking Ahead
AtScale has always believed semantic layers belong in the infrastructure layer, shared, governed, and reusable across the enterprise. We were the first to ship an independent, universal semantic layer over 13 years ago.
As OSI participants work to establish interchange standards, AtScale’s deep expertise in modernizing complex legacy environments provides practical insight into what semantic interoperability actually requires at enterprise scale.
Josh Klahr, Director of Analytics Product Management, captured the broader industry vision:
“Our collaboration with partners like AtScale establishes a unified, vendor-neutral standard for semantic data, ensuring clarity and consistency across the entire ecosystem. This initiative is essential for simplifying data operations, fostering innovation, and preparing organizations to build the next generation of AI applications.”
We are encouraged by the industry’s momentum toward openness and interoperability, and initiatives like OSI represent an important step in that direction. As these efforts evolve, AtScale is contributing real-world perspectives to help shape how open semantics function across BI, analytics, and AI.
Our focus is on providing practical insight, shaping approaches that handle real complexity, and helping ensure that open semantic infrastructure is not just interoperable in theory but also trustworthy and durable in practice. Collaboration across vendors, practitioners, and platforms is the only way this succeeds.
The future of enterprise analytics and AI depends on it.
Learn more about AtScale’s approach to open semantic infrastructure and how we are working with partners like Snowflake.
SHARE
2026 State of the Semantic Layer