Browse the documentation
Haomai Technology AI Brain

Core concepts in Haomai Technology AI Brain

One business question can involve figures in a table, rules in a document and permissions within an organization. Understanding the role of each helps explain what supports an answer and when a follow-up is needed.

Technical guide · November 20263 min read

Identify which source can answer the question

A data source holds queryable records, such as an order table or an approved business view. A knowledge base holds manuals, policies and explanatory documents. The ontology describes relationships between business objects, and metric definitions specify calculations. Each supplies a different kind of evidence.

For example, “How much of product A is available for sale this month?” calls for authorized inventory records. “How should we handle insufficient stock?” may call for a procedure. Finding the product manual does not establish the current stock level.

The concepts behind a question

ConceptQuestion it answersSynthetic example
Data sourceWhere are the underlying records?Order, inventory and collections views
Business objectWhat are we talking about?Product, organization, order
Metric definitionHow is the figure calculated?Value of confirmed orders, excluding canceled orders
Knowledge baseWhere do the rules and explanations come from?After-sales procedures and operating manuals
Tool or skillWhich tasks can be performed?Query data, retrieve documents, generate reports
PermissionsWhat can this identity see?Authorized records in this region only

From question to answer

Convert the objects, time range and metrics in the question into explicit conditions, then choose the appropriate query or retrieval path. Results should identify the source, reporting scope, citation location and reason for any missing data. A follow-up may retain the context or change the metric definition, so check the conditions again.

Three boundaries to keep in mind

  • Knowing a document exists does not grant the current user access to its contents.
  • An industry package still needs to be connected to data and pass business acceptance tests.
  • An analytical recommendation records a proposed action. Execution and business outcomes require separate evidence.
These are business and technical concepts. Object names, configuration settings and available tools depend on the deployed release.

Try “permissions,” “data source,” “manufacturing” or “API.”