Analysis
A query can run cleanly and still answer the wrong business question. Ask for "active customers" and you get a number; whether that is the number finance would sign is another matter.
A news item dated November 2, 2026, headlined "NVIDIA Releases Auto Ontology: An Open Semantic Layer That Grounds AI Agent SQL," reports that NVIDIA released Auto Ontology as an Apache 2.0 library. The excerpt says it generates a semantic layer from enterprise metadata, grounds agent SQL in certified definitions, and serves context through MCP, a command-line interface and an API. The full article was not read.
That puts definitions closer to where an agent forms its query, which could help keep terms like "active customer" consistent. But the excerpt offers no evidence that the generated layer is accurate, that agents give better answers with it, or that it works on real company data. A certified definition helps only if it matches the business's actual rules and stays current.
Before relying on a tool like this, ask who checks each generated definition, how corrections reach the source of truth, and what happens when a definition changes. Then put the same business question to it and compare the result with known records, reading the SQL as well as the answer.
Source: NVIDIA Releases Auto Ontology: An Open Semantic Layer That Grounds AI Agent SQL · 2026-11-02 · Excerpt only · full context unverified
Why it matters
Shared definitions only help if someone owns checking them, so test any such tool against records you already trust.