Natural-language query over SAP data — text-to-SQL patterns and limits

Two techniques wearing one label

"Ask your data a question in plain English" describes two structurally different mechanisms, and mixing them up in a proposal is how a delivery estimate goes wrong. The first, semantic-layer-mediated NLQ, is what SAC's Just Ask and, through it, Joule's Analytical Insights do: the question is matched against a pre-modelled analytic model — measures, hierarchies, labels already defined and reviewed — and the answer can only ever be as flexible as that model. The second, text-to-SQL, lets an LLM read table and column metadata from a governed catalog and generate its own query on the fly, closer to what Databricks Genie does against Unity Catalog: no pre-built model constrains the question, but nothing pre-built protects the answer either.

Why it matters

• "Natural-language query" is not one technique: semantic-layer-mediated NLQ and text-to-SQL fail in opposite ways — one fails visibly by silence, the other fails invisibly by plausibility — and scoping without naming which one leaves the real risk unaddressed.
• Access-control governance (catalog permissions) and semantic correctness (does the query mean what the business means) are different guarantees; text-to-SQL can deliver the first perfectly while still failing the second.
• Most business requirements for "ask your data anything" turn out, on inspection, to be requirements for trustworthy answers to a bounded set of questions — the semantic-layer-mediated case, not the open-ended one the phrase implies.

Key points

• Semantic-layer-mediated NLQ (Just Ask, Joule Analytical Insights): answers only through a pre-modelled, reviewed vocabulary; fails by silence or a traceable model gap.
• Text-to-SQL (Genie-style pattern): an LLM reads catalog metadata and writes its own query; fails by plausibility — a syntactically valid, confidently wrong answer.
• Governance (who can see what) rides on the catalog in both patterns and is well handled; semantic correctness (does the query mean what the business means) is structurally harder for text-to-SQL.
• The two are not mutually exclusive within one architecture — routine questions via semantic-layer NLQ, exploratory workloads via text-to-SQL over the same data products.
• A proposal naming only "natural-language query" without specifying the pattern has not scoped the skills, failure modes or governance story that differ between the two.
• Most real business requirements are for trustworthy answers to a bounded question set — the semantic-layer case — even when phrased as "ask anything."

Terms used on this page

Semantic-layer-mediated NLQ

Natural-language querying answered through a pre-modelled analytic model with governed measures, hierarchies and labels.

Text-to-SQL

An approach where an LLM generates and executes its own query against catalogued tables, without a pre-built semantic model.

Catalog-level governance

Access control enforced by the data catalog (e.g. Unity Catalog) regardless of which querying pattern is used above it.

Confidently wrong answer

A syntactically or fluently correct-looking answer that is semantically incorrect and gives no visible signal of the error.

Trusted asset (Genie)

A data-team-curated, verified question-to-SQL or metric mapping that a text-to-SQL system checks before falling back to free query generation.

GraphRAG

Retrieval that walks explicit graph relationships (SPARQL/RDF) rather than vector similarity or generated SQL — SAP HANA Cloud's Knowledge Graph Engine implements this for relationship-shaped questions neither NLQ pattern in this page answers well.

Principal propagation (MCP)

The mechanism carrying an MCP caller's real, signed-in identity through to an SAP ABAP user, so a text-to-SQL agent built over an MCP-exposed API inherits access control automatically.

Validation layer

A component built specifically to check a text-to-SQL system's generated queries against known-correct patterns before trusting the answer — the missing piece that catalog-level governance alone does not provide.

Sources

1. SAP News Center — SAP Business AI: Release Highlights Q1 2026 (base for the Just Ask/Analytical Insights semantic-layer path)
2. SAP Community — MCP Gateway in SAP Integration Suite
3. SAP Community — principal propagation for MCP servers, SAP Integration Suite to S/4HANA
4. SAP Community — querying RDF graphs with SAP HANA Cloud Knowledge Graph Engine (SPARQL as a structured alternative to text-to-SQL)
5. Databricks documentation — Genie (AI/BI) overview (fetched 2026-09-27)
6. Snowflake documentation — Cortex Analyst and the declared semantic model (fetched 2026-09-27)
7. SAP News — Autonomous Enterprise: AI Agent Hub and AI Governance Assistant (Sep 2026)
8. Spider 2.0: Evaluating Language Models on Real-World Enterprise Text-to-SQL Workflows — arXiv
9. BIRD: A Big Bench for Large-Scale Database Grounded Text-to-SQL — arXiv
10. A Semantic-Layer-Mediated Agent for Natural Language to SQL over Heterogeneous Enterprise Databases — arXiv
11. LinkAlign: Scalable Schema Linking for Large-Scale Multi-Database Text-to-SQL — arXiv
12. Analytical Insights in Joule — SAP Architecture Center
13. Joule with SAP Datasphere: step-by-step setup guide — SAP Community
14. Making NLQ More Explainable, Reliable and Actionable — SAP Community
15. Genie documentation — Databricks
16. Overview of semantic views (Cortex Analyst) — Snowflake documentation
17. SAP TechEd 2026 (Berlin & virtual, 27–29 Oct 2026) — session AS263: Powering autonomous agents with natural language data access