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
analysisProduction pattern

AvioBook’s airline data agents put identity and evidence ahead of autonomy

The proof of concept is notable less for what it automates than for what it refuses to: identity-scoped tools retrieve operational data, evidence accompanies each answer, and people retain the decision.

Airline data agents with guarded data path and human approval.
AI-generated diagram
By The News Desk· Sep 10, 2026the quick take — two AI hosts go live when you do

AvioBook’s new Connected Analytics design is a useful enterprise data-agent case study precisely because the agent is not the final authority. The system is still a proof of concept, but its boundaries are unusually concrete: two role-specific agents, tenant-scoped identity, a governed tool path into operational data, evidence returned with every answer, and human review before an operational judgment changes anything. AWS and AvioBook describe the architecture in a co-authored technical post.

One data plane, two jobs

AvioBook Connect already organizes airline operations around a live “flightroom” for each flight, combining API-generated events with crew and dispatcher messages. AvioBook’s own product page describes those rooms as the coordination surface for flight stakeholders.

Connected Analytics sits above that record. One agent is designed for airline managers doing historical analysis and reviewing delay codes. A second is designed for operations-control dispatchers asking about live disruptions and downstream effects. The separation matters: each user sees the agent authorized for that role, rather than one general assistant with every tool.

The request path is also explicit. A JSON Web Token carries the user identity into Amazon Bedrock AgentCore. The agent calls MCP-exposed tools through AgentCore Gateway; Lambda functions then query Amazon Athena. Athena resolves schema through AWS Glue Data Catalog and scans Parquet data in Amazon S3, partitioned by airline. That means the access boundary is enforced along the route to the data, not merely described in the prompt.

Evidence is part of the interface

The agents are designed to return the operational events supporting an answer. For delay-code review, the system compares a recorded code with the underlying sequence of events and flags inconsistencies as a second opinion. It does not automatically rewrite the code. AvioBook and AWS say outputs remain advisory, outside certified airworthiness systems, with an airline manager or dispatcher responsible for the decision.

This is the design lesson for other enterprise data teams: a trustworthy answer needs more than generated SQL or a fluent explanation. It needs a traceable identity, constrained tools, tenant isolation, retrieved evidence and a named human owner for consequential actions.

Do not confuse the baseline with an agent result

The post says an existing AvioBook Connect customer avoided more than 4,000 hours of delay over a summer season, including 124 hours attributed to reduced phone calls and manual coordination. Those are Connect baseline results, not measured benefits from Connected Analytics.

Likewise, the post’s estimate that a two-minute average turnaround reduction could be worth about $240,000 per month for a 200-flight-per-day carrier is explicitly illustrative. Connected Analytics has not yet demonstrated that saving in production. AvioBook says a focused week with AWS validated the proof-of-concept architecture and gave the team confidence to productize it; proactive anomaly alerting remains an area under investigation.

That distinction is the right one to preserve. The architecture is concrete enough to learn from. The business case is plausible enough to test. Neither should be reported as a shipped, measured agent outcome yet.

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.