Use AI Datasets to define the data context that Board AI Agents can use. A clear AI Dataset helps the Board AI Agent select the right data, explain results, and answer questions with the correct scope.
This article covers:
Dataset metadata for Board AI Agents
Use clear Dataset metadata to help Board AI Agents understand when to use an AI Dataset.
Write Dataset names and descriptions for business users, not only for Developers. Avoid technical names that users may not know.
A good Dataset description should explain:
The functional purpose of the Dataset.
The measures included in the Dataset.
The key dimensions used for analysis.
The main use cases supported by the Dataset.
The scope and level of detail.
The time period covered by the Dataset.
Avoid unclear or technical information, such as:
Acronyms without context.
Table names.
Cryptic Cube names.
ETL notes.
Internal technical comments.
Example
Dataset Name. P&L Actuals and Budget (Monthly)
Dataset Description. Income statement measures by Legal Entity, Cost Center, Account, and Month. Includes Actual, Budget, and Forecast versions. Use this Dataset to analyze financial performance over time, including revenue, costs, margins, and operating profit. This Dataset supports variance and driver analysis across periods, scenarios, and business dimensions. Business context for BoardAI Agents
Business context for Board AI Agents
Board AI Agents use Dataset metadata and Agent configuration to understand which Dataset to use and how to interpret the data.
Define the same business context in:
Dataset metadata, such as "Dataset Name" and "Dataset Description".
Board AI Agent configuration, such as Knowledge Base content, uploaded documents, or fine-tuning inputs.
The Dataset description should give a short, high-level summary of the data exposed in the Dataset.
When relevant, include:
The source system or data domain, such as ERP (Enterprise Resource Planning), CRM (Customer Relationship Management), or a custom source.
The high-level scope and purpose of the Dataset.
Key inclusions or exclusions, such as intercompany eliminations, adjustments, or Entity scope.
Currency context, such as local currency or group currency.
Data freshness, such as daily close, D+1, or intraday data.
Do not add detailed logic to the Dataset description. Keep mappings, normalization rules, and complex assumptions in Agent configuration, such as solution manuals or contextual files.
Data Configuration for Board AI Agents
Use clear data configuration so Board AI Agents can interpret the Dataset correctly.
Use clear Block headers
Use clear and consistent Block headers. Board AI Agents use Block headers to understand the meaning of each Block.
For example, use clear names such as:
Actual
Budget
Variance
Drivers
Unclear or inconsistent Block headers can cause the Board AI Agent to misinterpret the data or return incorrect explanations.
Align Blocks at the correct level of detail
Align Blocks so the Board AI Agent can compare and drill through data consistently.
Blocks do not need to use the exact same Cube design. However, the data exposed to the Board AI Agent must be comparable and clear for the target use case.
Make sure the Dataset structure supports:
Consistent comparisons.
Clear drill paths.
Correct interpretation for the target use case.
Use Cube versions for better performance
Use Cube versions when you need faster performance.
Aggregated Cube versions can:
Reduce query execution time.
Improve response time for top-level questions.
The "Axes" tab
Configure the "Axes" tab carefully. The Board AI Agent can use only the structures that are explicitly defined in the Quick Layout, including axes, axis settings, and drill paths.
The Board AI Agent cannot infer structures, drill paths, or selectable dimensions that are not defined in the Quick Layout.
Choose one main Entity in "By Column"
Use one main Entity in "By Column" to keep the view readable and consistent for analysis and commentary.
Limit Entities in "By Row"
Use no more than 2 or 3 Entities in "By Row".
Too many Entities can increase the Layout size and the number of combinations that the Board AI Agent must evaluate.
Use "Show all" only when needed
Disable "Show all" by default.
Enable "Show all" only for specific use cases, such as rule-based calculations.
.png)
The “Select” tab
Selections act as scope guardrails. The Board AI Agent can read these Selections, but cannot override them. This prevents the Board AI Agent from mixing scopes and helps keep analysis and commentary consistent.
Use Selections to define:
Entity scope, such as group only, region, or specific business units.
Scenario or version, such as Actual, Budget, or Forecast.
Currency or reporting view, such as local currency or group currency.
Inclusions or exclusions, such as excluding intercompany data or using an adjusted view.
.png)
Rules and governed Entities
If an Entity has rules applied, configure the Entity so the Board AI Agent can apply the rules at the correct level of detail.
Set the Entity as:
Mandatory. The Entity is always present in Board AI Agent queries.
Most nested. The rules are applied at the correct level of detail.
Additional best practices
Include only the data needed for the target use cases. This helps keep answers stable and relevant.
Expose only the Entities that the Board AI Agent must use in the Quick Layout.
Configure Entities clearly as "By Row", "By Column", "Mandatory", or "Nested". These settings control how the Board AI Agent can pivot and reconfigure the Layout when a user requests a different view.
Do not add Entities to the Quick Layout if they are already enforced as Dataset Selection filters. These Entities define a fixed scope and are not best used as pivot axes.
Data Attribute descriptions for Board AI Agents
Data Attribute descriptions help Board AI Agents understand your data in business terms, not only through technical names. Use descriptions when an Entity, Cube, KPI, or abbreviation is ambiguous, technical, or specific to a planning process.
A Developer or Power User can write a plain-language description directly in the Dataset configuration. The Agent can read the description when it encounters an ambiguous term and use the description to answer with the correct business context.
When to write Data Attribute descriptions
KPIs with non-obvious abbreviations, such as NCM, EBITDA, or CtA.
Entities or Cubes with business-specific definitions, such as available stock that excludes transit inventory.
Attributes where the name alone is ambiguous, technical, or specific to a use case or planning process.
When to skip Data Attribute descriptions
Standard time dimensions, such as Year, Month, or Period.
Self-explanatory Entities, such as Currency or Legal Entity.
Attributes that the Board AI Agent already interprets correctly.