Before starting to use AI capabilities in Board to analyze information from your Screens and Capsules, it is necessary to configure your Datasets and Agents.
Review the technical requirements described in Getting started with Board AI Agents.
This article covers the following topics:
How do the Agents interact with Board?
Agents do not query raw data directly. Instead, they interact with Board by retrieving data directly from Data Models through the Dataset abstraction layer to preserve performance, consistency, and security.
The interaction works as follows:
User request and context capture. The user asks a question or starts a task. The system captures the relevant context such as current page, selections, time range, and model scope.
Orchestration and task routing. The Orchestrator determines which Agent is best suited to the request and what steps are required to complete it reliably.
Data retrieval via the AI Semantic (Dataset) abstraction layer. Agents retrieve information through Datasets, which define selections and pivot structures. These requests are executed by the Board engine, not by the Agent itself.
Execution by the Board engine. The Board engine performs aggregations and calculations, applying the model logic and enforcing security. This ensures results match the “single source of truth” used across Board.
Reasoning, explanation, and guidance. The Agent interprets the results, explains drivers and assumptions, and proposes next steps or actions appropriate for the workflow.
This separation is deliberate: the engine remains responsible for deterministic calculation and governance, while Agents focus on interpretation, guidance, and productivity.
How to create and configure a Dataset
Before starting to use AI capabilities in Board to analyze information from your Screens and Capsules, it is necessary to configure your Datasets and Agents.
Learn more in About Datasets, or follow the steps to Create a Dataset in Board.
When creating or editing a Dataset, make sure to enable the “Dataset for AI” option so it can be used by your Agents.
How to create and configure an AI Agent
After creating your Dataset, you will be able to create and configure your Agents as follows.
Inside a Data Model, the “AI Agent Configuration” tile is situated under “Analytics”. By clicking on this tile you will reach the AI Agents area from where you can create, edit, set as default and delete Agents.
.png)
.png)
To create and configure an AI Agent, follow these steps:
Click the plus icon
.png)
Select which Board Agent you would like to configure.
Only the AI Agents configured in your License will display.
FP&A Agent. Focuses on Financial Planning and Analysis with capabilities related to Balance Sheet analysis, cash flow insights and Profit and Loss review.
Controller Agent (requires the FCR solution installed). Brings intelligence and automation to financial close and consolidation by surfacing true exceptions, validating financial statements and cash flows, and reducing manual reconciliation.
Click CONFIGURE
The “Overview” Tab
Name your Agent
Associate one or more Datasets to your Agent (either an existing Dataset from the dropdown menu or a new Dataset by clicking on the “external link” icon and being redirected to the Datasets Catalog)
(optional) Agent Tools. This area will show you all the plugins that can be used by your Agent.
(optional) Upload .txt files from your device to add solution guidance, use-case instructions, and customer-specific business logic. These files are used for model fine-tuning.
These files guide the Agent’s behavior and definitions but they don’t change the underlying Datasets or security.
Use a .txt file when you need to:
- Standardize definitions (KPIs, sign conventions, solution specific lingo)
- Explain business rules / calculations (allocations, rule-based KPIs)
- Provide mappings & terminology (acronyms, synonyms → Board entity names)
- Add process guidance (reconciliation approach, navigation)
Tips
- Include keywords & synonyms users will type (improves matching for the Agent to look up and retrieve): file name, short description of what the file contains at the start of the file and its intended usage by the agent. This guides the Agent to look into the file content based on user prompt.
- Keep guidance aligned to what is available in linked Datasets and Quick Layout
- Upload a maximum of 5 files (UI indicates file count/size limits)
The “Suggested Prompts” Tab
The Suggested Prompts feature enables Developers and Power Users to create repeatable, optimized analyses that serve as a guided starting point for end users. It allows for the standardization of common use cases, ensuring consistency and efficiency in how analyses are performed.
To create custom suggested prompts for your AI Agent to display in a Screen, continue configuring:
Click on “ADD SEGMENT”
Choose a name and icon to associate with your Segment, and click “CREATE”
When a Segment is defined you can create specific prompts by clicking on the plus icon in front of your Segment’s name. After that you can specify the prompt’s title and description to instruct the Agent on specific analyses (this information is not displayed to the End User)
Notice the “Preview” area on the right side of the popup window which shows how your Agent will be displayed to the End User.
Currently, Board allows the creation of 6 segments with 6 prompts each.
Click “CREATE” to complete your Agent’s setup.
Once the Agent is configured in the Data Model, you can use it from your Screens and Capsules by clicking on the AI icon .png){kind=small}
situated on the Top Menu. Note that this icon will only be visible once an Agent is linked to the Screen.
Learn more about how to interact with your Agents in a Screen.