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
guideDeployment guide

An analytics answer needs a receipt—not just a screenshot

Decision provenance links the question, chosen meaning, analytical plan, evidence and reviewer disposition without pretending the result was permanently correct.

Final answer versus decision receipt.
AI-generated illustration
By The News Desk· Oct 4, 2026the quick take — two AI hosts go live when you do

A screenshot preserves an answer. It does not preserve why a team believed the answer.

That gap matters when an analytics agent turns an ambiguous business question into a number that later informs a decision. Spotonix defines decision provenance as the trace connecting the original question to its interpretation, analytical plan, data and logic, answer, and human review. The useful part of the definition is that it reaches beyond ordinary data lineage: lineage can show where data moved, while decision provenance records the business meaning and human disposition attached to its use.

The minimum receipt

Spotonix’s worked example starts with the question, “Why did enterprise revenue miss plan?” A reviewable record would preserve four layers:

  1. The question as asked. Do not replace it later with a cleaner prompt that makes the answer look more inevitable.
  2. The selected meaning. Record which definitions of “enterprise” and “revenue” were used.
  3. The analytical path. Keep the comparison window, grain, logic, sources and result together.
  4. The human disposition. Record whether the reviewer accepted, corrected or escalated the result.

This turns review into more than checking generated SQL. A reviewer can recover the original request, inspect the assumptions that shaped it, follow the sources and logic, and see what happened after the answer was presented.

Preserve corrections without rewriting history

The reviewer disposition is especially important. If a person corrects the enterprise-customer definition, the system should not silently overwrite the first interpretation and present the revised answer as though it had always been obvious. Keeping both the original path and the correction shows where the analytical process failed and what changed.

That record can also prevent a past answer from becoming accidental policy. Spotonix explicitly says provenance should show what was known and accepted at the time without treating the prior decision as permanent truth. Definitions, source data and business conditions can all change; a reusable answer still needs to expose its date and assumptions.

What the receipt cannot prove

Decision provenance is not an accuracy certificate. Spotonix’s own boundary is clear: provenance supports inspection and accountability, but does not prove the underlying decision was correct. A complete record can faithfully document a bad definition, faulty source or mistaken approval.

The practical deployment test is therefore modest but useful: ask whether a later reviewer can recover the original question and selected meaning; whether the plan, sources, logic and result remain linked; and whether human acceptance, correction or escalation is recorded without erasing the earlier state. If an analytics agent can only export a chart or final number, it has preserved the output—not the decision path.

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.