Skip to main content
Version: v12

Brinqa Attribute Assistant — User Guide

Version: 1.0 Last Updated: April 9, 2026


1. What Is the Attribute Assistant?​

The Attribute Assistant is Brinqa's AI-powered feature that learns patterns in your data and predicts missing attribute values automatically. You train a model by selecting a target attribute (the field you want the AI to fill in) and a set of input attributes (the fields the AI uses to make its predictions). Once trained, the model can be run against your data to fill in gaps — and every AI-predicted value is clearly flagged so you always know what came from the AI versus your source systems.


2. Key Concepts​

Target attribute — the attribute whose missing values you want the AI to predict (e.g., severity, priority, category).

Input attributes — the attributes the AI uses as signals to predict the target. You can select these manually or let the system recommend them with Auto Select.

Auto Select — an automated process that analyzes your data and recommends the most useful input attributes for your chosen target. It also shows each attribute's influence on the prediction.

Test run — a trial run that uses a partial sample of your data. Results are not saved and do not affect your live data. Use this to validate performance before committing.

Train — a full training run on your complete dataset. This produces the model that will be used for inference. After training the influence of the selected features will be available.

Inference — the process of running a trained model against your data to generate predictions for missing attribute values.


3. Training a Model​

You can train a model in two ways: using Auto Select (recommended for most users) or by manually selecting your input attributes.

Auto Select analyzes your data and recommends the input attributes that are most predictive of your target. This is the easiest way to get started.

  1. Navigate to a data model. Open the data model you want to work with — for example, CVE Record.

  2. Open the AI Attribution Agent tab. Click the AI Attribution Agent tab within the data model view.

  3. Create a new model. Click the + icon button on the right side of the panel.

  4. Select a target attribute. Choose any valid target attribute from the dropdown — this is the field you want the AI to predict.

  5. Click Create.

  6. Run Auto Select. Click Auto Select Attributes. The system will begin analyzing your data in the background.

  7. Wait for results. You must stay on the page while the auto selection is running. This may take a few minutes. When it finishes, you will see the recommended input attributes along with their influence scores (how much each attribute contributes to the prediction), and an accuracy that shows how well we expect this model to predict your target attribute.

  8. Activate the model. If you are satisfied with the recommended attributes, check the Active checkbox and click Update in the lower-left corner.

  9. Choose Test or Train. You will be taken to an overview page showing all models for this data model. From here:

    • Select Test to do a trial run using partial data (results are not saved).
    • Select Train to do a full training run on your complete dataset. This will generate the model used in inferencing later.
    • Note: Until the model is trained, it cannot be used for inferencing.

3.2 Option B: Training with Manual Attribute Selection​

If you have domain knowledge about which attributes are most relevant, you can select them yourself instead of using Auto Select.

  1. Navigate to a data model. Open the data model you want to work with — for example, CVE Record.

  2. Open the AI Attribution Agent tab. Click the AI Attribution Agent tab within the data model view.

  3. Create a new model. Click the + icon button on the right side of the panel.

  4. Select a target attribute. Choose any valid target attribute from the dropdown.

  5. Click Create.

  6. Select input attributes. Manually choose between 3 and 12 attributes that you believe will be useful for predicting the target. Select attributes that are likely to have a meaningful relationship with the target value.

  7. Activate the model. Check the Active checkbox and click Update.

  8. Choose Test or Train. You will be taken to an overview page showing all models for this data model. From here:

    • Select Test to do a trial run using partial data (results are not saved).
    • Select Train to do a full training run on your complete dataset. This will generate the model used in inferencing later.

4. Running Inference​

Once you have a trained model, you can run inference to generate predictions across your data. Inference is triggered through a Flow, and runs all of the available trained models.

  1. Open Settings. Click the gear/settings icon in the top-right corner of the Brinqa interface.

  2. Navigate to Flows. Under Automations, click Flows.

  3. Find the inference flow. Click View All beside Recent Flows or Recent Folders, then type AI Attribute Inferencing Flow into the search bar.

  4. Open the flow. Click on the AI Attribute Inferencing Flow search result.

  5. Launch. Click Launch. A confirmation popup will appear — click the Launch button on the popup without filling in any fields.

The flow will run in the background. When it finishes, your data will be updated with AI-predicted values. All predictions are flagged as AI-generated and will require your approval before being used to train future models.


5. Troubleshooting​

Training failed or produced poor results​

The most common cause is insufficient data. The Attribute Assistant requires at least 1,000 rows with complete data across all selected input attributes to train effectively.

If you used Auto Select, be aware that the most statistically useful attributes may not always have the best data coverage. For example, Auto Select might identify a combination of attributes that is highly predictive, but only a small number of rows have values for all of those attributes simultaneously. If the overlap falls below 1,000 complete rows, training will fail.

What to try:

  • Check your data coverage. Review whether the selected input attributes have enough populated rows. Look for attributes with high null rates that may be reducing the usable row count.
  • Switch to manual selection. If Auto Select chose attributes with sparse data, try manually selecting attributes that you know have broad coverage across your dataset, even if they are slightly less predictive individually.
  • Reduce the number of input attributes. Each additional attribute you add tightens the requirement that a row must have all values filled in. Using fewer, well-populated attributes may give the model more rows to learn from.
  • Add more data. If possible, enrich your data sources to improve coverage for the attributes you want to use.

Model is trained but predictions seem inaccurate​

  • Run a Test first. Always start with a test run to evaluate model quality before committing to a full training run.
  • Review the input attributes. The attributes you selected may not have a strong relationship with the target. Consider trying different combinations or using Auto Select to see which attributes the system recommends.
  • Check for data quality issues. Inconsistent, outdated, or incorrectly formatted values in your input attributes can reduce prediction quality.

6. Quick Reference​

TaskWhere to Go
Create a new modelData Model → AI Attribution Agent tab → + button
Auto select attributesInside model → Auto Select Attributes
Activate a modelInside model → check Active → Update
Test a modelModel overview page → Test
Train a modelModel overview page → Train
Run inferenceSettings → Automations → Flows → search AI Attribute Inferencing Flow → Launch

For additional support, contact your Brinqa account representative or visit the Brinqa help center.