What was announced
Within about two weeks of each other in June, Databricks and Snowflake both announced context layers for AI agents, Snowflake with Cortex Sense, Databricks with Genie Ontology.
Both are aimed at the same problem: an agent sitting on top of a data warehouse doesn't inherently know what your business means by "revenue," which table is authoritative, or how your org structures its data. Without that shared understanding, every query becomes an exploration from scratch. Snowflake reports that Cortex Sense raised accuracy on structured context tasks from 24% to 86% by giving agents that grounding upfront that's Snowflake's own reported figure from its preview, not an independently verified benchmark, worth noting since Cortex Sense is still in private preview with no GA date announced.
The two approaches work differently. Genie Ontology builds a persistent knowledge graph and ranks definitions using what Databricks calls "ontorank" a PageRank-inspired mechanism that weighs creator credibility, usage breadth, and recency, drawing on Databricks data plus more than 50 connected apps like Slack and SharePoint. It's also in preview, reporting 84.5% accuracy on its own benchmark. Cortex Sense works more at query time, it reads from Snowflake's governed definition store and reorders what's relevant the moment a question arrives, drawing on query history and BI dashboards.
My take
Watching both of these ship within two weeks of each other tells me something: semantic grounding is becoming as important to these platforms as raw query performance used to be. That's a real shift. A few years ago the pitch was "how fast can we run your SQL." Now it's "how well do we understand what your SQL means." I think that trend continues, and I'd expect every major platform to have some version of this within a year or two.
What I don't think gets solved soon is the cross-vendor version of it. Genie Ontology and Cortex Sense are both, by design, answers within one platform's walls. Nobody's incentivized to build the version that spans competitors; Databricks isn't going to build a context layer that makes Snowflake's data equally trustworthy, and vice versa. If your business runs both, you're on your own for bridging that gap, at least for now.
NexusIQ doesn't have an ontology or a ranked business-glossary layer; that's a different, harder problem than what I've built. What it does have is a governance layer that travels with the query regardless of which engine it hits: RBAC and row-level security enforced consistently across Databricks, Snowflake, MySQL, and Exasol, plus RAG retrieval over governed documents with the same role-based filtering. That's not the same as "semantic grounding," but it's the same instinct, don't let the platform boundary be where your control ends. You can see how that's wired together on the Nexusiq Architecure Page
Platforms will keep building smarter context inside their own walls. The harder, more valuable problem is making that context travel across walls and almost nobody is building for that yet.
Source: Snowflake and Databricks Summits 2026: What Actually Matters — PointFive
Additional reference: What Is Genie Ontology? — Typedef