Custom AI Models

Prev Next

Create and train custom AI Models to return better results when generating Load and Truck Postings.

Train new models based on customer email requests and user-submitted requests from the Text to Load and Text to Truck Posting Lightning Web Components (LWCs).

Create a Custom AI Model

Custom AI Models are generated from existing gpt-4.1 and gpt-4.1-mini models and trained on returned Load and Truck Posting records from the AI. Currently, a custom AI Model cannot be trained multiple times. A new model must be created and trained off of an existing model, including custom models.

  1. From the AI Console, choose the Model Training icon to view the training console.

  2. Click the plus icon in the AI Models section.

  3. Enter a Model Name and search for the Parent Model. Typically, the Parent Model is gpt-4.1 or gpt-4.1-mini.

    1. The gpt-4.1 model typically returns better results, but takes more time to finish. The gpt-4.1-mini returns results more quickly, but may not be as accurate.

  4. Check the Enabled box. Only enabled models may be used.

  5. Check the Is Custom Model box.

    1. All user-created models are custom models.

  6. Click Save.

Create Training Sets

Create AI Model Training Sets to train the AI Model. Training Sets consist of AI Model Training records.

Training records are created from the output generated by AI. Review the output and correct any fields before creating training records.

Create a Training Set record by clicking the Plus icon to open the modal.

In the modal, enter a name and an optional description for the AI Model Training Set.

Return to the AI Console.

Review and correct Load or Truck Posting records as necessary to train a custom AI Model.

  1. Select a Load or Truck Posting record from the AI Console activity section.

  2. View the input sent to the AI to generate the record.

  3. In the output form AI section, view returned field values.

    1. Update any values that returned incorrectly.

  4. Click the Save button to keep any field changes.

  5. View the JSON payload returned from the AI.

Click the Create Training Record button to open the modal.

Select the Training Set to add the Training Record to it. Click the Create button.

Create additional Training Records and add them to the Training Set.

Training Sets should consist of approximately 10 to 200 Training Records.

When the AI Model Training Set is selected, the records are displayed in the Console below.

View Training Set information with the Edit button.

The Edit Training Set modal allows for changing the Training Set name or adding a description. The available buttons complete the following actions.

  • Preview

  • Upload

  • Save

  • Clone

  • Cancel

View the information included in the Training Set with the Preview button. Each Training record added to the Training Set is displayed in the preview. If any records require changes, return to the individual record and update fields as necessary.

Only “correct” Load or Truck Posting records should be used in Training Sets.

Click the Upload button to send the Training Set.

Click Save to keep any changes and return to the AI Console. Select Clone to create a copy of an existing Training Set. Choose Cancel to discard any changes and return to the AI Console.

Conduct Training Sessions

To use a Training Set, it must first be sent to OpenAI. Click the Upload button. The Upload Status appears in the AI Console as Uploaded when the Training Set is successfully received.

A single Training Set may be used to train multiple AI Models.

When the Training Set is uploaded, conduct a Training Session.

Create a new AI Model Training Session.

Select the AI Model Training Set and a Validation Training Set. The Training Set makes changes to the AI Model and the Validation Training Set verifies that those changes are correct.

A Validation Training Set is not required but is recommended.

Click Save to create the AI Model Training Session record and return to the AI Console.

Select the AI Model to train. The AI Model Training Session appears on the Console. Click the Edit icon to view the record.

Click the Start Training button.

View the current Status of the Training Session from the AI Console.

Use the Refresh button on the AI Console to view the latest status update.

Training an AI Model may take several minutes.

When the Training Session completes successfully, the newly-trained AI Model may be used to create Load and Truck Posting records.

  1. Return to the Activity screen of the AI Console.

  2. Click the Configuration button in the banner.

  3. Select the new model from the AI Model picklist.

  4. Click Save.