live wire
▸Snowflake agent evals can pin judge versions; unpinned metrics may roll forward and shift scoresSnowflake▸ThoughtSpot makes Spotter 3 default for new instances; it requires download rights and always shares query data with an LLMThoughtSpot documentation▸MotherDuck adds image classification and object location to prompt_jev(), returning typed SQL outputs with confidence scoresMotherDuck▸BigQuery conversational analytics adds Montréal and Toronto; mixed-region agents still default to globalGoogle Cloud Documentation▸Fabric enables Azure OpenAI Copilot access by default, but OpenAI-subprocessor and cross-region processing/storage switches remain offMicrosoft Learn▸Snowflake Cortex Agents now continue by default when named tools are inaccessible, emitting warnings instead of failing the runSnowflake documentation▸Databricks Genie One can use group memberships to tailor answers, while Unity Catalog remains the access-control boundaryDatabricks documentation▸Fabric data agents can query Power BI models with Read alone; Build and workspace access are unnecessaryMicrosoft Learn▸Fabric data-agent Git configs expose column flags, but selecting a table still includes every columnMicrosoft Learn▸Google Data Agent Kit offers remote Spark kernels, but notebook exports can still move results into local Pandas or ArrowGoogle Cloud Documentation▸Cube says production analytics agents need explicit partial-result states, caller-filtered model search and durable artifactsCube▸Thinking Machines names Bridgewater behind ReViSQL-K2.6, citing 92.97% SC-16 accuracy at 15% of frontier-model costThinking Machines Lab▸AWS says a mismatched data/tool/prompt release cut its Field Sales agent pass rate from 78% to 38% in pre-productionAWS Artificial Intelligence Blog▸NL2SQL Multi-Agent ships evaluation artifacts across BIRD, CSpider, Archer and KDD DataSpace—but no software licenseGitHub▸Snowflake agent evals can pin judge versions; unpinned metrics may roll forward and shift scoresSnowflake▸ThoughtSpot makes Spotter 3 default for new instances; it requires download rights and always shares query data with an LLMThoughtSpot documentation▸MotherDuck adds image classification and object location to prompt_jev(), returning typed SQL outputs with confidence scoresMotherDuck▸BigQuery conversational analytics adds Montréal and Toronto; mixed-region agents still default to globalGoogle Cloud Documentation▸Fabric enables Azure OpenAI Copilot access by default, but OpenAI-subprocessor and cross-region processing/storage switches remain offMicrosoft Learn▸Snowflake Cortex Agents now continue by default when named tools are inaccessible, emitting warnings instead of failing the runSnowflake documentation▸Databricks Genie One can use group memberships to tailor answers, while Unity Catalog remains the access-control boundaryDatabricks documentation▸Fabric data agents can query Power BI models with Read alone; Build and workspace access are unnecessaryMicrosoft Learn▸Fabric data-agent Git configs expose column flags, but selecting a table still includes every columnMicrosoft Learn▸Google Data Agent Kit offers remote Spark kernels, but notebook exports can still move results into local Pandas or ArrowGoogle Cloud Documentation▸Cube says production analytics agents need explicit partial-result states, caller-filtered model search and durable artifactsCube▸Thinking Machines names Bridgewater behind ReViSQL-K2.6, citing 92.97% SC-16 accuracy at 15% of frontier-model costThinking Machines Lab▸AWS says a mismatched data/tool/prompt release cut its Field Sales agent pass rate from 78% to 38% in pre-productionAWS Artificial Intelligence Blog▸NL2SQL Multi-Agent ships evaluation artifacts across BIRD, CSpider, Archer and KDD DataSpace—but no software licenseGitHub
nl2sql.ai
guideMigration guide

Oracle’s Select AI matrix makes 26ai an accuracy upgrade—not an NL2SQL prerequisite

19c retains core natural-language querying and agent tools, while 26ai adds feedback, automatic schema selection and vector-backed retrieval.

Oracle 19c versus 26ai capabilities compared.
AI-generated illustration
By The News Desk· Oct 15, 2026the quick take — two AI hosts go live when you do

The upgrade question is narrower than it looks

Oracle’s current Select AI capability matrix shows that natural-language-to-SQL is available in Autonomous AI Database 19c and Oracle AI Database 19.30, as well as the newer 26ai lines. The same matrix also marks AI agents, conversations, SQL agent tools and custom PL/SQL tools as available on those 19c editions.

That means teams do not need to move to 26ai merely to keep a basic in-database NL2SQL or agent workflow. Oracle’s own release comparison describes 19c as retaining the common foundation: chat, NL2SQL, synthetic-data generation, agents, summarization and translation.

What 26ai actually unlocks

The practical upgrade boundary is the context and correction layer around generation. Oracle marks NL2SQL feedback and automatic object selection as available in Autonomous AI Database 26ai and Oracle AI Database 23.26.1, but not in the compared 19c editions. It also limits Select AI’s RAG capability and vector-index controls to those 26ai lines in this matrix.

Automatic object selection is especially consequential for large schemas. Oracle says it builds a vector index over objects in an AI profile, then uses semantic similarity to choose relevant metadata for each prompt. Instead of sending every table definition to the model, the database can narrow the schema context before generation. Feedback provides a separate route for incorporating user corrections and preferences into later NL2SQL behavior.

The agent boundary follows the same pattern. SQL, web-search, notification and custom PL/SQL tools remain listed for 19c, but the built-in RAG tool is marked 26ai-only. A team can therefore keep action-oriented database agents on 19c while lacking Oracle’s integrated vector-backed retrieval path.

A migration test, not a feature checklist

For practitioners, the matrix suggests three tests before upgrading: measure whether schema size is overwhelming the current object list; determine whether corrected queries need a supported feedback loop; and identify any workflow that depends on Select AI-managed vector indexes or RAG. If none applies, 19c still has the core query and agent surfaces documented by Oracle.

One detail deserves careful reading: the package table lists a FEEDBACK procedure on 19c, while the top-level capability summary marks NL2SQL Feedback unavailable there. Teams should validate the exact feedback behavior they need rather than treating a procedure name as proof of feature parity. Oracle’s matrix is useful precisely because release compatibility is granular: the command surface and the complete product capability are not always the same thing.

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