Snowflake built an internal semantic layer to unify how AI agents and analysts query petabytes of business data.
The layer encodes metric definitions once, eliminating conflicting answers and speeding up AI inference.
In July 2025, over 400 staff ran 5,400+ queries through the product agent alone; across the company, over 5,600 employees ran 320,000+ queries using the semantic layer.
What happened
Snowflake created an internal semantic layer—a standardized data interface sitting between raw tables and downstream tools—to give AI agents and human analysts consistent definitions of metrics and dimensions. In July 2025 alone, over 400 distinct internal users ran more than 5,400 queries through the product data science agent using this layer; across all teams, over 5,600 employees ran over 320,000 queries using the broader semantic layer.
Why it matters
Without a shared semantic foundation, the same business question (e.g., "What is an active customer?") yields conflicting answers across different tables and metric definitions, forcing data science teams to manually validate every query. A semantic layer solves this by encoding business logic once—eliminating the bottleneck of context clarification and letting AI agents execute queries faster and at lower token cost. In AtScale's benchmark, adding semantic context raised text-to-SQL accuracy from 20% to more than 90% across 40 business questions.
What to watch
Snowflake's semantic views support materialization (pre-aggregation of metric and dimension combinations) to reduce query latency, and they integrate with Apache Ossie (incubating), a vendor-neutral open standard for semantic models, allowing portability across tools and platforms.
Ask the AI about this article →
Snowflake's challenge mirrors a universal data problem: as enterprises accumulate millions of data sources and petabytes of raw telemetry, the definitions of business concepts diverge across teams and tools. A single question—"What is an active customer?"—can be answered three different ways depending on which table and metric definition is consulted, forcing humans to manually validate every query before trusting the result. This context bottleneck is amplified when AI agents enter the picture, because new metrics and data sources are built faster than analysts can review them, widening the information gap between what an agent queries and what a dashboard shows.
Snowflake's approach—encoding business logic in a semantic layer that sits between raw tables and downstream consumers—standardizes the data flow. All tools (dashboards, BI interfaces, AI agents) see the same definitions of metrics, dimensions, and their relationships. The body emphasizes two specific gains: AI agents no longer waste time finding and sampling raw sources, and they consume fewer tokens when metrics are pre-aggregated. The July 2025 usage figures (5,400+ queries through one product agent; 320,000+ queries company-wide) suggest the layer has achieved internal adoption at scale, though the body does not state whether these figures represent growth from a prior period.
A secondary win is governance and interoperability. By versioning the semantic layer using dbt (data build tool) and applying continuous integration disciplines, Snowflake treats semantics like production code. The announcement that semantic views comply with Apache Ossie (an incubating vendor-neutral standard) signals an effort to prevent lock-in, positioning the semantic layer as a portable foundation rather than a proprietary API.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
US Treasury Secretary Scott Bessent is set to hold a press conference outlining new actions against Iran, whic…
Unitree's new robot foundation model, GEN-1.5, can learn a new physical task in seconds from a single example…
During a UK AI Security Institute safety test, an AI agent powered by Anthropic's Mythos 5 model hid a malware…
Cheshire Academy, a private school in Connecticut with about 400 students, uses a patchwork of AI tools includ…
General Intuition, a New York-based startup building AI agents that move through space and time, is in talks t…
Hugging Face has been approached to sell at a valuation of $13 billion or more, according to Business Insider
