Energy / Delivered project

AI Chat Interface for Energy Data

An agentic query layer that translates business questions into governed filters, queries, maps, and cited analytical responses.

Explore the page ↓

Geospatial energy data connected through a visual intelligence layer — Energy · Data → Intelligence → ActionPhoto by Luke Chesser · Unsplash License ↗

01 / Problem statementDelivered project · Energy

The operating problem

Valuable operational and geospatial datasets remain inaccessible to many users because the schema, tools, and query language require specialist knowledge.

Conditions the system must survive03 operating constraints

Operational and geospatial schemas require specialist knowledge

Natural-language questions can be ambiguous or unsafe to execute

Permissions must apply to rows, fields, tools, and generated answers

02 / How we approached it

From operating uncertainty to testable evidence.

The work was decomposed into four engineering decisions. Each one produced an artifact the customer could inspect, test, and carry into deployment.

01Frame

Create a semantic layer over approved datasets

Evidence produced

Approved semantic layer and permission model

02Evaluate

Translate questions into inspectable query plans

Evidence produced

Question-to-query benchmark across analytical task types

03Engineer

Enforce row, field, and tool permissions

Evidence produced

Execution safety and ambiguity review with analysts

04Operationalize

Return results with filters, assumptions, and sources

Evidence produced

Cited answer interface with visible filters and query trace

03 / Deployment record

Deployed around the workflow—not beside it.

The system boundary includes where inference runs, how evidence reaches existing tools, and how people handle uncertainty after launch.

Implementation statusDelivered project

Topology

Private agent and semantic-query services placed behind enterprise data and identity controls.

Integration

SQL, time-series, GIS, asset, and document sources expose approved semantic entities and scoped tools.

Operation

Ambiguous questions trigger clarification; query plans are validated before execution and answers retain source and permission context.

04 / Evaluation metrics

What must be measured before the system earns trust.

Evaluation covers model behavior, workflow burden, and production performance. The metric defines the gate; the customer baseline and acceptance threshold define the target.

01Evaluation gate

Semantic parse accuracy

How it is measured

Correct entities, filters, time windows, and intended operation.

What it decides

Determines when a plan can execute or needs clarification.

02Evaluation gate

Execution success

How it is measured

Valid, permission-compliant queries returning the intended result shape.

What it decides

Measures tool and schema reliability.

03Evaluation gate

Answer faithfulness

How it is measured

Claims supported by returned data, filters, and cited sources.

What it decides

Controls whether narrative answers can be trusted.

04Evaluation gate

Time to evidence

How it is measured

Time from question to analyst-accepted result.

What it decides

Quantifies accessibility improvement for complex data.

05 / Customer perspective

Value has to appear in the customer’s operating day.

What matters in practice

The assistant should show the interpreted question, filters, query plan, assumptions, and sources before its narrative answer.

01Observable value signalBroader access to complex data

02Observable value signalShorter path from question to evidence

03Observable value signalReusable analytical workflows

04Observable value signalAuditable query execution

06 / Technology context

Tools follow the system—not the other way around.

Final architecture depends on data quality, operating conditions, integrations, risk, and evaluation criteria established during discovery.

LLMsAI agentsSQLGISSemantic layer

Test the operating assumption

Define the evidence required to move from possibility to production.

Discuss this use case ↗

AI Chat Interface for Energy Data

Valuable operational and geospatial datasets remain inaccessible to many users because the schema, tools, and query language require specialist knowledge.

From operating uncertainty to testable evidence.

Create a semantic layer over approved datasets

Translate questions into inspectable query plans

Enforce row, field, and tool permissions

Return results with filters, assumptions, and sources

Deployed around the workflow—not beside it.

What must be measured before the system earns trust.

Semantic parse accuracy

Execution success

Answer faithfulness

Time to evidence

Value has to appear in the customer’s operating day.

Tools follow the system—not the other way around.

Related operational systems.

Load forecasting and asset anomaly intelligence

Automated Energy Data Intelligence Dashboard

Define the evidence required to move from possibility to production.