Third-Party AI Form Prefill

Follow

About AI Form Prefill

AI Form Prefill lets you complete a third-party form from information you already have. You describe the details in your own words, upload documents, or do both. AI then fills the empty fields on the form, and you review each suggestion before the form is submitted.

The suggestions help you complete the form. They do not replace your review, and you decide what is submitted.

 

Before You Start

You see the prefill card at the top of a form when three things are in place: your organization has AI enabled, your administrator has turned on AI Form Prefill in System Administration > AI Features (see Enabling AI Features), and prefill has been enabled on that form. If you can open a form but don't see the card, contact your administrator. Your administrator can request prefill on a form through a support request.

 

Add Information

  1. Open a form that has AI Form Prefill enabled. The card at the top reads Complete this form using information you already have.

  2. Select Add information to open the Add information window.

  3. Describe the details in your own words in the text box, upload one or more documents, or do both. To upload, select the upload area or drag files onto it.

  4. Submit your information using the Complete Form button.

Supported document formats are PDF, docx, doc, txt, xlsx and xls. You can upload up to 5 files, and each file can be up to 4.5 MB. A file over the size limit is not added, and a message names the file and says it is larger than 4.5 MB and was not added.

When you provide a description and documents together, they are used together. Your description can add details the documents don't hold, correct an error in a document, or say which sections of the form to fill. If two different values are found for the same field, then neither will be used unless one is more specific than the other.

Stay on the form while your information is processed. If you leave the form before processing finishes, the suggestions are not applied.

 

Which Fields Are Filled

Prefill fills only fields that are empty. A field that already holds a value is not changed. Hidden fields and read-only fields are not filled.

Prefill supports text box, text area, email, number, date and select fields. Field type names are configurable per form, so the names on your forms can differ. For a select field, the suggestion is always one of the options available on that field. User picker fields are not filled.

Prefill works on forms with more than one page. A field that appears only when a condition is met is filled when its condition is satisfied.

 

Review The Suggestions

After processing, the card changes to Suggestions have been applied. Each field that was filled is highlighted and shows an AI icon and an Accept button.

b0166cc0-773d-4d09-9c39-bbfda9f0209a.png

For each suggested field, do one of the following:

  • Select Accept to keep the suggestion.

  • Edit the value. Your edited value replaces the suggestion and the field is no longer waiting for review.

  • Reject the suggestion by clearing the field. If the field is required, you still need to enter a value before you continue.

You must resolve every suggested field on the current page before you can continue or save the form.

To fill more fields, select Add more information and provide more details or documents. Only fields that are still empty are filled.

 

Check The Suggestions Against Your Source

AI-generated suggestions may contain inaccuracies. Before you accept a suggestion, check it against the description or document it came from. Take particular care with text area fields, where the suggested wording may be composed by the model rather than copied from your source, and with select fields, where the model chooses between options.

 

Your Data

Suggestions are generated by Claude Sonnet 5.5, a model from Anthropic, which GAN Integrity accesses through Amazon Bedrock. Your data is not used to train the model.

Your data is processed in the same region as your GAN Integrity system. If your system is hosted in the EU, the model processes your data in AWS regions in the EU. If your system is hosted in North America, the model processes your data in AWS regions in the United States.

Uploaded documents are not kept, and no text is extracted from them. They are sent to the model in full. An audit record keeps your description, the names of your files, the form fields sent to the model with their current values, and the suggestions. The audit record is retained.

Was this article helpful?
0 out of 0 found this helpful