live wire
IBM makes watsonx Orchestrate AgentOps, custom LLM judging and Bedrock-agent discovery generally availableIBMSchemaGate 0.1.45 fixes broken Oracle ADB wallet connections and an OCI stack pinned 28 releases behindSchemaGatePDI’s Amazon Quick procurement agent grounds spend answers in vendor, category and contract contextAWS Business Intelligence BlogBigQuery’s ML.METRICS example returns 0.84 accuracy but 0.30 macro-F1 on the same 100-row classification queryGoogle Cloud BigQuery docsSchemaGate 0.1.44 auto-selects sentence embeddings, lifting bundled-schema retrieval from 90/98 to 93/98SchemaGateSchemaGate 0.1.43 adds read-only SQL execution with per-principal table checks—and documents unauthenticated client assertionsSchemaGateDatabox adds reusable AI Analyst Skills with personal/company scope, auto-matching and marketplace installsDataboxFabric previews an AI builder for data-agent instructions, source guidance and example queriesMicrosoft FabricDatabricks trains data-agent retriever to stop early or spend bounded extra search steps, reporting 5.8-second latencyDatabricksThoughtSpot adds SpotterCode coding agent to its Visual Embed PlaygroundThoughtSpotLongMemEval-S audit: 67–73% of restore-fixable 80k-budget errors came from evicted evidence under three policiesarXivSchemaGate 0.1.42 adds dimension-aware retrieval and fixes complex multi-table SQL promptsSchemaGateSnowflake agent toolsets can silently drop inherited tools when callers lack accessSnowflake DocumentationLooker’s VS Code extension reaches GA with MCP-assisted LookML generation, editing and validationGoogle Cloud Looker release docsIBM makes watsonx Orchestrate AgentOps, custom LLM judging and Bedrock-agent discovery generally availableIBMSchemaGate 0.1.45 fixes broken Oracle ADB wallet connections and an OCI stack pinned 28 releases behindSchemaGatePDI’s Amazon Quick procurement agent grounds spend answers in vendor, category and contract contextAWS Business Intelligence BlogBigQuery’s ML.METRICS example returns 0.84 accuracy but 0.30 macro-F1 on the same 100-row classification queryGoogle Cloud BigQuery docsSchemaGate 0.1.44 auto-selects sentence embeddings, lifting bundled-schema retrieval from 90/98 to 93/98SchemaGateSchemaGate 0.1.43 adds read-only SQL execution with per-principal table checks—and documents unauthenticated client assertionsSchemaGateDatabox adds reusable AI Analyst Skills with personal/company scope, auto-matching and marketplace installsDataboxFabric previews an AI builder for data-agent instructions, source guidance and example queriesMicrosoft FabricDatabricks trains data-agent retriever to stop early or spend bounded extra search steps, reporting 5.8-second latencyDatabricksThoughtSpot adds SpotterCode coding agent to its Visual Embed PlaygroundThoughtSpotLongMemEval-S audit: 67–73% of restore-fixable 80k-budget errors came from evicted evidence under three policiesarXivSchemaGate 0.1.42 adds dimension-aware retrieval and fixes complex multi-table SQL promptsSchemaGateSnowflake agent toolsets can silently drop inherited tools when callers lack accessSnowflake DocumentationLooker’s VS Code extension reaches GA with MCP-assisted LookML generation, editing and validationGoogle Cloud Looker release docs
nl2sql.ai
guideDEPLOYMENT GUIDE

Put business rules in profiles before data agents reach dashboards

Oracle and Looker’s current guidance points to a concrete rollout order: separate context by tool, keep environment permissions local, then widen the user surface.

Business rules in profiles; access stays local before dashboards open.
Side by side: what changed
By The News Desk· Sep 10, 2026the quick take — two AI hosts go live when you do

A useful data-agent rollout does not begin with a chat box. It begins by deciding where business meaning, data access and agent behavior live—and keeping those responsibilities separate.

Oracle’s reusable business-context guide makes that separation concrete. It puts SQL-specific definitions—such as how to interpret “last week,” what counts as a delayed flight and whether cancelled flights are excluded—in an NL2SQL profile. Retrieval constraints belong in a separate RAG profile. Agent persona and orchestration duties normally belong in the agent role and task definitions.

Put durable rules at the layer that enforces them

This is more than prompt organization. Oracle says profile-level role and additional_instructions are added to each augmented prompt that uses the profile. The same SQL definitions can therefore apply across RUNSQL, SHOWSQL, EXPLAINSQL and related actions without every application repeating them.

Oracle also documents a call-level override through DBMS_CLOUD_AI.GENERATE. An application can temporarily change date handling, output shape or role for one request without mutating the shared profile. That yields a useful operating distinction: persistent business definitions go in the profile; request-specific presentation and scope stay with the call.

The same boundary should survive deployment. Oracle’s portable agent-team guide says team definitions can move as Open Agent Specification JSON or canonical PL/SQL. AI profile definitions, custom tool code, manifests and test assets can travel with them. Credential values, network ACLs, privileges, endpoints and target data-access policy remain responsibilities of the destination environment.

That means portability is not permission inheritance. A production promotion should rebind and validate local controls rather than assume that a working development agent carries authority with it. Oracle recommends recreating prerequisites first, then testing team objects, tool functions and representative requests after deployment.

Expand the conversational surface last

Looker’s release notes show the user-facing layer widening while configuration boundaries remain visible. Conversational Analytics verified queries are generally available, dashboard data agents remain in preview, copied dashboards preserve agent instructions, configurations and sources, and self-service models are limited to their assigned user database connections.

Taken together, the two vendors suggest a defensible rollout checklist:

  1. Encode durable business definitions and approved objects in the relevant SQL or retrieval profile.
  2. Keep SQL, RAG and orchestration instructions separate instead of relying on one oversized prompt.
  3. Treat agent definitions as versioned release artifacts, but recreate credentials and privileges in each target environment.
  4. Validate tool behavior and representative questions before promotion.
  5. Only then expose the agent through dashboards or wider user groups.

This sequence cannot guarantee correct answers. It does make meaning, portability and access boundaries inspectable before the interface reaches more users—a stronger production gate than a polished demo response.

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.