live wire
▸Oracle Select AI generates SQL/PGQ for property graphs—but errors if one AI profile mixes graphs with tables or viewsOracle▸Domo’s data team standardizes MCP builds with shared scaffolding; beta toolkits inherit user permissionsDomo▸Databricks proposes five outcome signals for agentic BI—and says dashboard counts and license usage miss decision qualityDatabricks▸One vibe-coded database app needed 56+ human test hours and three repair rounds despite AI Studio’s verification claimsMancas & Mancas / arXiv▸BigQuery AI.IF’s default cost mode trains a local model for most rows—and drops row-error thresholdsGoogle Cloud▸AWS’s multi-tenant data-agent pattern gives the broker Lambda no data permissions; Cognito roles carry Lake Formation scopeAWS Big Data Blog▸Amazon OpenSearch’s optimized engine adds nested agent traces, but AWS reserves it for log analytics—not vector searchAWS▸Two interpretation-based provenance semantics cannot express SQL bag difference, new theorem showsKolaitis / arXiv▸Databricks Feature Views reuse one definition for training and live scoring, but online lookup still returns latest available stateDatabricks▸Databricks Genie’s Slack and Teams previews default channel answers to public, even when viewers lack source-data accessDatabricks▸Snowflake’s two-tool GraphRAG agent beat a seven-tool graph setup 70% to 60% on 22 ontology questionsSnowflake Engineering▸Semantix opens an Apache-2.0 semantic-first data-agent stack, but labels the project WIP and has no release or adoption yetGitHub / icosmos-space▸Confluent defines Streamhouse as shared current state for agents and apps, but says broader semantics and governance are still to comeConfluent▸Graph-repair retrieval cut F1 by 0.126 on one suite; a learned router used fewer tokens than always-on retrieval in all sixLin et al. / arXiv▸Oracle Select AI generates SQL/PGQ for property graphs—but errors if one AI profile mixes graphs with tables or viewsOracle▸Domo’s data team standardizes MCP builds with shared scaffolding; beta toolkits inherit user permissionsDomo▸Databricks proposes five outcome signals for agentic BI—and says dashboard counts and license usage miss decision qualityDatabricks▸One vibe-coded database app needed 56+ human test hours and three repair rounds despite AI Studio’s verification claimsMancas & Mancas / arXiv▸BigQuery AI.IF’s default cost mode trains a local model for most rows—and drops row-error thresholdsGoogle Cloud▸AWS’s multi-tenant data-agent pattern gives the broker Lambda no data permissions; Cognito roles carry Lake Formation scopeAWS Big Data Blog▸Amazon OpenSearch’s optimized engine adds nested agent traces, but AWS reserves it for log analytics—not vector searchAWS▸Two interpretation-based provenance semantics cannot express SQL bag difference, new theorem showsKolaitis / arXiv▸Databricks Feature Views reuse one definition for training and live scoring, but online lookup still returns latest available stateDatabricks▸Databricks Genie’s Slack and Teams previews default channel answers to public, even when viewers lack source-data accessDatabricks▸Snowflake’s two-tool GraphRAG agent beat a seven-tool graph setup 70% to 60% on 22 ontology questionsSnowflake Engineering▸Semantix opens an Apache-2.0 semantic-first data-agent stack, but labels the project WIP and has no release or adoption yetGitHub / icosmos-space▸Confluent defines Streamhouse as shared current state for agents and apps, but says broader semantics and governance are still to comeConfluent▸Graph-repair retrieval cut F1 by 0.126 on one suite; a learned router used fewer tokens than always-on retrieval in all sixLin et al. / arXiv
nl2sql.ai
analysisPattern analysis

The semantic layer is becoming an agent’s executable contract—not just a BI catalog

Four enterprise implementations converge on the same design: business meaning must be machine-readable, queryable and tied to security before an agent can act safely.

By The News Desk· Nov 7, 2026the quick take — two AI hosts go live when you do

Enterprise data agents are turning the semantic layer from documentation for analysts into an input that software actively executes. Recent first-party designs from Google Cloud, Databricks, Tableau and TS Imagine differ in scope, but they converge on one practical idea: an agent needs machine-readable business meaning between the user’s words and the underlying data.

Meaning is moving into the query path

Google Cloud’s preview QueryData service makes the pattern explicit. Its context sets combine pre-validated question-and-SQL blueprints, reusable SQL predicates called facets, and value-search rules that map phrases such as “Bay Area” to stored values. QueryData uses those objects during value resolution, disambiguation and SQL synthesis; parameterized secure views can then restrict execution to rows available to the authenticated application user. Google Cloud documentation

Databricks is encoding a narrower but related contract in Unity Catalog. Beta relationships between metric views record direction, cardinality and join conditions so AI assistants and BI tools do not have to infer joins from column names. The relationships remain metadata rather than enforced joins, and Databricks warns that tools must choose among ambiguous paths and cycles. That makes the contract executable by clients without making it automatically correct. Databricks documentation

The contract is broader than SQL

Tableau Knowledge extends the idea into a graph built from published data sources, workbooks, dashboards and metrics. Tableau says subject-matter experts can curate inferred relationships, while Tableau Agent and external MCP clients retrieve the resulting graph. Connections to Snowflake, Databricks, Workday and other outside sources are still described as coming soon, and the product is currently packaged in Tableau Cloud+ and Tableau+. Tableau

TS Imagine applies the same principle to a regulated trading workflow. Its TSIQ architecture places an ontology and semantic layer between probabilistic models and deterministic trading, risk and certified-data systems. The company’s example is average daily volume: market venue, time window and liquidity type must be fixed before the number is useful. TS Imagine says generated orders can be traced to the initiating conversation and context, although its post provides architecture claims rather than an independent error-rate evaluation. TS Imagine

What teams should build

The shared lesson is not that one vendor has solved semantics. It is that a glossary alone is no longer enough. A production contract should identify canonical query patterns, allowed join paths, value-resolution rules, metric parameters and the authorization boundary applied at execution. It should also preserve which version of that contract produced an answer or action.

The limitations matter: Google’s service is preview, Databricks relationships are beta and non-enforcing, Tableau’s external connectors are forthcoming, and TS Imagine does not publish independent performance evidence. Still, the direction is consistent across four different products. The semantic layer is becoming code that shapes what an agent may ask, how it constructs the query and which result it is allowed to return.

sources

Filed by The News Desk. Corrections: desk@nl2sql.ai · Our standards →

comments · 0

    Comments are moderated before they appear. Your email is used once to confirm it is you — never shown, never sold. Corrections and questions get an answer from the desk when we have one.

    more from the desk