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
newsEnterprise agents

IBM puts Dun & Bradstreet’s commercial graph behind an MCP server

The watsonx Orchestrate catalog listing is a useful example of enterprise data agents moving from bespoke connectors toward governed, reusable context services.

IBM-style commercial graph service piping verified company data into an orchestration layer.
AI-generated illustration
By The News Desk· Aug 26, 2026the quick take — two AI hosts go live when you do

IBM has added the Dun & Bradstreet Commercial Graph to the watsonx Orchestrate Agent Catalog through a Model Context Protocol server, giving enterprise agents a packaged route to verified company data. IBM’s Aug. 12 announcement says the connector is live now.

The immediate use cases are not generic chatbot demos. IBM points to finance, credit, procurement, compliance, sales and marketing workflows, with the D-U-N-S Number acting as a common company identifier across systems of record. The MCP server can retrieve and analyze D&B commercial data while operating inside D&B’s existing quality controls, permissions and audit trails, according to the announcement.

Why this matters for data-agent teams

For practitioners building natural-language interfaces to enterprise data, generating a plausible query is only one part of the system. The agent also needs authoritative entity context, current data and a controlled way to reach it. IBM’s catalog model packages that access as a reusable capability rather than asking each implementation team to build and maintain another proprietary connector.

That separation is important. A text-to-SQL layer can answer questions about tables already inside a warehouse, while a commercial-graph service supplies verified external context such as company identity, financial-health indicators and relationship mappings. An orchestration layer can combine the two without pretending they are the same data source or applying the same access policy to both.

IBM says watsonx Orchestrate provides centralized control for building, monitoring and scaling agents across cloud and on-premises systems. The company also says the D&B server can support near-real-time company research, supplier-risk checks, credit assessment and cross-system workflows.

What to verify before adoption

The announcement makes broad deployment claims, including that teams can reduce agent-to-data connection work from months to days, but it does not publish a benchmark or a detailed pricing model. Buyers should therefore treat that statement as vendor guidance, not a guaranteed implementation timeline.

A practical evaluation should test four things: which D&B fields the server exposes; how watsonx and D&B permissions combine; what audit records are available for each agent call; and whether the returned company identity can be joined reliably to internal customer, supplier and account records. Teams should also measure freshness and failure behavior for the exact jurisdictions and business entities they use.

The larger signal is architectural: enterprise agents are beginning to consume governed data products through standardized tool interfaces. For NL2SQL systems, that creates a cleaner boundary between query generation, business semantics and external enrichment—and a clearer checklist for deciding which layer is responsible when an answer is wrong.

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