Copilot Auto Extract fields

There are two ways that you can use Copilot field extraction.

1. Manually configure Copilot fields in bulk

This Copilot extraction method uses a large language model alongside natural language instructions to extract data. Manually provide simple descriptions in bulk for each piece of data to extract. There is no need for additional configuration, scripts, or training.

Use this method of Copilot field extraction when a document contains only a few pieces of data that you want to extract. The time it takes to configure a few fields may be quicker than using Copilot proposed fields and excluding all but those few fields.

You can also extract inferred information that is not printed on the document. See Inferred or derived fields for more information.

2. Use Copilot proposed fields to automatically add fields in bulk

This Copilot extraction method sends a training document to the Copilot engine that returns a rated list of proposed fields, complete with names, data types, location, and descriptions. Select which proposed Copilot fields to use.

Use this method of Copilot field extraction when a document has a lot of data that you want to extract.

Copilot Auto Extract works best for Western languages. The results for other languages, such as Chinese or right-to-left languages, are not guaranteed.

For both Copilot extraction methods, a maximum of 100 pages is sent to the Copilot engine. A maximum of 50 characters is allowed for field names and a maximum of 300 characters is allowed for descriptions.

You can combine auto extract fields with other field extraction methods such as suggested fields, lassoed fields, AI tables, and detected tables.

The Copilot Auto Extract method is disabled for predefined document types, and test documents.