Respect access scope
Enforce role, tenant, row and column permissions in the data service. Understanding a question does not grant a model more access.
Identity → authorized viewsData Intelligence API
Data, semantics and business reasoningLet data
answer business.
A business question deserves an answer with evidence.
Translate business language into an explicit analysis plan. Compute within authorized data, then return the metrics, findings and sources together.Follow an analysisAn embedded capability for data platforms, business systems and agentsDiscuss integration
Beyond the number.
Understand what supports it.
Growth, fulfilment and margin require different analyses. Explore how each question changes the metric definition, visualization and evidence behind the answer.
Start with a business questionInteractive illustration · synthetic data
Choose a question
The plan for this question
- Metric
net_revenue- Grain
- Region × month
- Scope
- Posted revenue · one currency · cancelled records excluded
How much did revenue grow in September, and which region contributed most?
Month-on-month revenue growth+13.6%
East contributed 62.5% of the revenue increase.
September revenue was CNY 2.68m versus CNY 2.36m in August, an increase of CNY 0.32m. East added CNY 0.20m, the largest contribution.
AugustSeptember
East+20
100
120
South+6
76
82
North+6
60
66
Result evidence3 regions · 2 complete months
(268 − 236) / 236 = 13.6%This is a contribution analysis; it does not establish the causes of growth.
Inspect the query and underlying data
Illustrative query · no database connection
SELECT region, month, SUM(net_revenue) AS revenue
FROM authorized_revenue_view
WHERE month IN ('2026-08', '2026-09')
AND posting_status = 'posted'
GROUP BY region, month;| Region | August revenue | September revenue |
|---|---|---|
| East | 100 | 120 |
| South | 76 | 82 |
| North | 60 | 66 |
finance.revenue_by_regionNext, break down East by product and compare price and volume movements.
Business semantics
Agree on the definition.
Then calculate the answer.
“Revenue” can mean contracted value, invoiced value or recognized revenue. Entities, metrics, time and business rules need a shared definition.Illustrative metric contract
Net revenue
net_revenuev1EntityRecognized revenue records
revenue_entryAggregationSum of net amounts
SUM(net_amount)Time dimensionRevenue recognition date
recognized_atDimensionsRegion, product and customer
region · product · customerFiltersExclude cancelled records
status = 'posted'Access scopeInherit identity and tenant
viewer → allowed_rowsOne definition serves questions, charts and analysis tasks. Join grain, deduplication and currency belong in the contract to avoid fan-out and inconsistent calculations.
Understand.
Compute.
Explain with evidence.
ID Axis organizes analysis into traceable tasks. Models interpret and plan; tools calculate; validation determines which findings can enter the answer.
Explore ID Axis- 01
Interpret the intent
Identify metrics, time, dimensions and comparison logic. Clarify missing definitions before substituting assumptions for business rules.
Question → intent object - 02
Build a semantic plan
Choose a path through available metrics and entities. Check relationship cardinality, aggregation grain and filters.
Semantic model → reviewable plan - 03
Execute within boundaries
Run read-only analysis through authorized views and restricted credentials. Bound scan cost, time and output size; retain execution status.
Plan → computed result - 04
Validate and explain
Check empty results, zero denominators, duplicate joins and period completeness. Keep calculated facts, interpretations and next steps distinct.
Result → finding + source
The boundary matters
as much as the answer.
Generating SQL is only one part of enterprise analysis. Identity, query execution, data quality and evidence each need their own controls.
Bound execution
Combine read-only transactions, restricted accounts and query checks. Writes and external actions belong in a separate approval and execution flow.
Read-only → budget → auditSurface data issues
Expose missing values, inconsistent grain and incomplete periods. When evidence is insufficient, return that limitation explicitly.
Quality checks → applicable scopeRetain the evidence
Record metric version, query scope and result sources. Correlation is not causation; a suggested diagnosis is not a verified fact.
Calculated facts → interpretation → open questionsDifferent operations.
Different questions.
Business performance
What makes up the growth?
Decompose revenue and margin by region, channel and product. Align period comparisons and contributions, then trace movements to specific segments.
Supply and fulfilment
Where did the delivery pattern change?
Connect orders, inventory and event times. Use consistent promise and completion definitions to locate delays and guide investigation.
Customers and service
Which signals deserve another look?
Use authorized transaction and service records to examine repeat purchases, activity and service changes. Preserve sample scope rather than generalizing from a few cases.
An API inside your business workflow
Keep the answer
where the work happens.
Embed data intelligence in a platform, workspace or agent. The host manages identity and data access; analysis returns presentable, traceable results. Your workflow remains in charge of business actions.
Your business systemIdentity · authorized data · questions
Question and context
Data Intelligence API
Semantic planning / bounded computation / validation
Results and evidence
Structured resultsMetrics and chartsSources and scope
Integration is assessed against your data environment. This demo does not connect to live data.
Technical references
Public references on semantic modeling, execution boundaries and data validation.
These references explain general engineering principles, rather than a specific implementation or measured performance. All demonstrated data and analysis are synthetic; capabilities and delivery boundaries are established during integration assessment.