Set up a use case

Create a use case to define the information to extract from a document for processing.

Before you begin

Several predefined use cases are available within their defined workflow areas. Check whether the available use cases meet your requirements before creating a new one. For more information, see Content Understanding integrations.

Role required: DocIntel Admin [sn_docintel.admin] or DocIntel Manager [sn_docintel.manager]

About this task

In a use case, specify the document type to process, the fields and tables to detect, the questions to ask, and where to store the results.

After you define a use case, you can begin processing documents for it in the related workflows.

Procedure

1. Navigate to All > AI Admin Hub > Skills.
2. In the workflow list, select Platform.
3. From the Platform skills list, search for Extract information from documents skill.
4. Select the options menu (), and then select Edit.
5. Select New use case.
6. Define a use case.

   
   
   

   1. Enter a name for the use case.
   2. Select a target table to store the document processing results for the use case.
   3. Select the language of the files to process for the use case.

      If the files contain multiple languages, select the primary language.

      For more information, see Languages supported by Content Understanding.

   4. Select a large language model (LLM) that will make predictions for the documents processed with this use case.

      For more information, see Large language models used by Content Understanding.

   5. Optional: Turn on image mode to process images more efficiently.

      Image mode sends pages to the LLM as images to use the visual capability of the multimodal LLM and any of the languages supported by it.

      Note:

      Selecting image mode reduces the page count limit to 10 pages per file.

      The image mode option is available when a multimodal LLM is selected.

      Note:

      Now LLM Service is a text-only model and doesn't support image mode.
   6. Select Save and continue.
7. Define information fields, questions, or tables.

   
   
   

   1. Select Add a field.
   2. Select the type of information to extract from the document.

      You can choose one of the following:

      Field

      Fields are used to extract a single piece of information in the document. For example, a document number or a customer name.

      Question

      Define a question to ask about the document.

      Table

      Tables are used to extract lists or tables of information. A table can have multiple columns. The number of list items or table rows doesn’t have to be known in advance.

      A field form displays based on the information type you selected.
   3. On the form, fill in the fields.
      The type of form depends on the type of field:

      • Field form fields
      • Question form fields
      • Table form fields
   4. Select Save.
      The field, question, or table is added to the Information list for the use case.
   5. Select Save and Continue.

What to do next

1. Test a use case
2. Add integration
3. Review and activate