How to Use Advocate’s AI Document Extraction Feature

Advocate’s platform empowers users with cutting-edge AI document extraction technology, enabling faster and more accurate insurance compliance workflows. This feature is designed to intelligently read insurance documents, extract relevant data, and streamline checklist completion — all while maintaining a layer of human oversight to ensure data accuracy.

Advocate’s AI Document Extraction feature uses machine learning models trained to read and interpret specific insurance documents. Once uploaded and associated with the appropriate policy, the system automatically extracts key data points and maps them to the relevant fields in the review checklist.

This process combines the speed and scale of AI with the precision of human verification, ensuring that all extracted data is reviewed for correctness.

Advocate’s AI extraction is currently supported for the following insurance forms:

• ACORD 28

• ACORD 25

As Advocate continues to develop and train its AI models, more document types will be supported over time.

You can initiate AI Document Extraction during several common workflows in the platform:

• Uploading a New Document and Associating it with a New Policy

  • After associating the document with the relevant policy, the system will auto-populate mapped fields with extracted data.

• Updating a Revised Document for an Existing Policy

Note: The AI extraction tool cannot be used to populate fields during the initial policy creation form. It activates after the document has been associated with a policy.

To enable AI extraction, ensure the relevant toggle is switched on during upload.

• A progress indicator shows the extraction status.

• Once complete, the UI will refresh to display extracted data in the checklist.

The AI Document Extraction process includes a guided three-stage review to ensure all extracted and required data is properly validated:

Once the extraction is complete and associated with a new policy, a green notification banner appears at the top of the checklist.

This stage presents all new data extracted by the AI that requires user confirmation. Each extracted data point is annotated in purple on the source document, helping users trace where the information was pulled from. This visual aid enhances accuracy and transparency in the validation process.

After reviewing and confirming the AI-suggested values, users can save the checklist and move on to reviewing the remaining non-AI fields.

This stage is triggered primarily when a revised or newer version of a document is uploaded to an existing policy.

A banner at the top of the checklist will be highlighted in orange. The AI identifies changes in the new document that conflict with previously saved inputs in the checklist. These conflicts require manual review and user confirmation to ensure the revised data reflects an actual update in coverage or terms.

This step ensures that only valid, intentional changes are accepted into the policy record.

After addressing AI-extracted entries and any conflicts, users must complete a final pass through the checklist.

This stage includes fields not populated by AI—either because the data was not available in the document, or because it pertains to information outside the scope of the AI-supported document types.

These fields require manual input before the checklist can be finalized.

This powerful feature reduces manual entry, enhances accuracy, and accelerates the compliance process—all while keeping humans in the loop to ensure data quality and compliance integrity.