Your data engineers have spent months getting your metric definitions right: revenue recognized the way finance approved it, churn calculated the way your exec team aligned on it, and pipeline logic that your rev ops team actually agrees on. And then a new tool arrives, and someone has to do it all again.
That duplication carries two costs. The first is time: many hours of work recreating structure, definitions, and relationships that already exist. The second is risk: every time a metric gets redefined in a new tool, there’s a chance it gets defined differently. This is where data governance breaks down, your AI agents start reasoning from inconsistently applied logic and definitions, and your teams end up with different results from each other and your agents, without understanding why.
And that’s assuming your analytics and AI are working from a real semantic layer at all. Tools that query tables directly let agents generate SQL against raw schemas, which means every answer is only as good as the model’s best guess at what your data means. Without a semantic layer, there’s no place for “revenue” to be defined, no logic to audit, and no way to know if the answer is right until someone checks it manually.
ThoughtSpot and Snowflake are solving this together. Available now in Early Access, ThoughtSpot natively imports Snowflake Semantic Views directly into Spotter Semantics—ThoughtSpot’s Agentic Semantic Layer. If your data team has already defined facts, metrics, dimensions, and relationships in a Snowflake Semantic View, that context flows into ThoughtSpot automatically. But they don’t just pass through.
Spotter Semantics takes what you define in Snowflake, enriches it with further metadata, and makes it operational in the context where your users are consuming insights—across AI agents (ThoughtSpot’s or custom), multi-agent pipelines, search, Liveboards, embedded analytics, and other applications.
ThoughtSpot layers its own reasoning on top, so that AI agents like Spotter interpret natural language questions from business users, navigate ambiguity, and connect concepts across your data model. Snowflake Semantic Views define business metrics—the formulae, the dimensions, the relationships—and can carry natural language context like synonyms and custom instructions. Spotter Semantics is where those definitions meet how your business actually talks about data, powering every AI and BI surface area your business runs on.
"The combination of Snowflake Semantic Views and ThoughtSpot’s Spotter Semantics represents a massive leap forward for data governance and analytics. Organizations can now define their semantic context for AI and BI once in Snowflake and use ThoughtSpot as a trusted intelligent context layer for AI and agents to scale those definitions across the enterprise. <br><br>This helps ensure that agentic workflows and AI-generated insights are based on a single, governed source of truth."
Why the Semantic Layer Matters for AI and Agents
When ThoughtSpot—or any agent or application running on ThoughtSpot—answers a question, it reasons from a semantic layer: a layer of context that tells it what your data means, how entities relate, and which business logic applies. Semantics also help make an answer auditable—the logic behind it traces back to verified definitions your team owns, not inferred from raw schemas.
A general-purpose AI tool or a BI tool without a mature semantic layer might look at raw tables and guess how to calculate "Net Revenue” for your specific use case. Sometimes, it’ll get close, but “close” doesn’t cut it. When it gets it wrong, you’re left with a hallucination that looks like a real answer with no lineage to trace back through.
With Spotter Semantics, now grounded in context from Snowflake Semantic Views, the agent is working from the actual logic your team owns. Think of it as the Rosetta Stone for enterprise data: a translation layer that lets agents interpret your data the way your business actually defines it.
This logic extends to your entire AI ecosystem. Whether you are using Spotter or orchestrating multi-agent pipelines, an agent’s reliability is capped by its context. Importing Snowflake Semantic Views into Spotter Semantics—and in the future, natively enriching your Snowflake Semantic Views back with insights from ThoughtSpot—means your AI, analytics, and agents are grounded in one governed context layer from the moment they're deployed.
This integration is also part of ThoughtSpot’s broader commitment to open semantic interoperability as a founding partner of the Open Semantic Interchange (OSI)—an open standard, championed by Snowflake and co-developed with other industry leaders, to ensure semantic definitions can move freely across the data ecosystem rather than staying trapped inside any single tool.
How to Get Started Today
Path 1: Import Snowflake Semantic Views with ThoughtSpot’s Native Integration
For teams that have already invested in building Semantic Views inside Snowflake, this path turns that investment into richer context for every ThoughtSpot experience immediately.
Here, we demonstrate how you can import a Snowflake Semantic View using ThoughtSpot’s native connector, save it as a ThoughtSpot Model, and ask your dashboard agent SpotterViz to build a new Liveboard using insights from your new model—all in just a few minutes.
Step 1: Create a Semantic View in Snowflake
Start by creating a Semantic View in Snowflake. Use this guide to get started.
Step 2: Import the Semantic View in ThoughtSpot in Seconds
In ThoughtSpot, go to your Data Workspace, and click Semantic Integrations. Connect to the Semantic View you wish to import through your existing Snowflake connection, and simply click to import within seconds. You have the option to review and edit formulas, before saving the view as a new ThoughtSpot Model.
Step 3: Get and Share Insights
Open Spotter or any other ThoughtSpot agent like SpotterViz, and ask a question on top of your new ThoughtSpot Model. ThoughtSpot uses the definitions from your Snowflake Semantic View as part of its reasoning. Share your answers from Spotter or publish Liveboards, so your team can go from insights to action.
Path 2: Bi-Directional Sync Using ThoughtSpot Agent Skills
ThoughtSpot’s agent skills can be called via API, embedded in applications, and orchestrated as a node inside multi-agent pipelines. This gives teams who prefer to work in AI-native workflows the option to perform bi-directional syncs between their ThoughtSpot Models and Snowflake Semantic Views today. These agent skills are available for use today.
Here, we demonstrate how you can leverage ThoughtSpot’s agent skills to move a ThoughtSpot Model into Cortex Code, create a Semantic View from the ThoughtSpot Model, and then push the Semantic View back into ThoughtSpot. No matter where your users consume insights, you can assure they are producing structured, explainable responses grounded in the semantic context your team has built.
One Context Layer, Everywhere It Matters
Your data team has already done the hard work of defining what your metrics mean. For teams governing that logic in Snowflake, the next step is straightforward: bring those definitions into Spotter Semantics, where ThoughtSpot enriches them with AI-native context and puts them to work across your agents, analytics, and every experience your business runs on.
The native integration launching today is just the foundation. We’re working towards making the native integration between Snowflake Semantic Views and Spotter Semantics fully bidirectional, as well as supporting automatic model refresh. For any teams who wish to use these capabilities today, ThoughtSpot agent skills are open source and available today for use in coding agents, including Cortex Code.
Request a demo today to see how you can build a context layer for your AI and agents—governed, enriched, and operational everywhere it matters.
