Over four and a half years at Snowflake, I watched enterprises run into a problem that’s about to get much worse: every major platform has built its own semantic layer, and none of them talk to each other.
Many platforms solve half the battle by getting data centralized and accessible. But as enterprises deploy more AI use cases, the real value shows up when that data actually means something consistent: the same definition of revenue, the same definition of customer, no matter who or what is asking.
That gap is about to matter a lot more than it used to. Every platform under the sun has built their own semantic layer, and frontier models from Anthropic and OpenAI expect you to supply that context yourself. None of these layers talk to each other.
Plenty of companies run Snowflake and Databricks side by side, use Power BI and Tableau, and deploy agents from Anthropic alongside models from OpenAI. Each platform carries its own metric definitions and context, and the result is inconsistent answers and agents that hallucinate.
For most of the last decade, that inconsistency wasn’t urgent. Semantic layers were a BI optimization, a way to speed up self-service reporting and cut down on duplicate metrics. AI changed that. A person querying a dashboard can tolerate some inconsistency; an LLM driving business decisions can’t. A hallucinating agent costs money, and conflicting metric definitions across agents erode trust fast.
There’s no real AI strategy without a data strategy, and no legitimate data strategy without correct context and understanding of that data. Enterprises need a semantic layer that works across platforms, not just within one.
The pattern I keep hearing
In nearly every conversation, I hear some version of the same thing: “We have semantic layers in all our platforms. The problem is they’re isolated from each other.”
When I bring up Open Semantic Interchange (OSI), a standard for portable, interoperable semantic definitions, people get it immediately. Enterprises want autonomy: a universal semantic layer that gives consistent context across every BI tool, data platform, LLM, and agent in the business. That’s genuinely valuable, and it’s why semantic layers are moving from a BI feature to core infrastructure for enterprise AI.
Why I joined AtScale
I’m here to help build the most critical region during this inflection point, the kind of work that compounds as we grow. I’m genuinely excited to help shape what AtScale becomes, and just as excited to recruit and develop A-players who want to build something special alongside us.
Three things drove the decision.
People. This isn’t a team of career passengers, it’s a group of genuine A-player drivers and builders, several who’ve been here since the early days and several who’ve joined recently at exactly the right inflection point. That combination of deep conviction plus fresh and hungry talent is rare and very special.
Timing. Google searches for “semantic layer” have climbed sharply over the past two years, and the market has shifted from treating it as a BI convenience to treating it as essential AI infrastructure. AtScale has been building this for more than a decade. The market is finally catching up to why it matters.
Technology. The moat isn’t the semantic layer concept that every platform has bolted one on. It’s that AtScale’s is universal and portable: one governed definition of a metric, consumed identically by any BI tool, any warehouse, any AI agent. Competitors are building walled gardens. AtScale built the layer that works across all of them, and that’s a decade-long head start that doesn’t get replicated with feature release.
What’s next
I’m now leading sales for AtScale in the East US, bringing the GTM and operations discipline I’ve learned over the years to the problem of universal semantic layers. This is a problem enterprises actually need solved, and for companies running multiple data platforms while deploying AI at scale, the timing couldn’t be better.
SHARE
How to Evaluate Context Platforms