Using prompt tags in generative AI prompts
- Updated: 2025/08/01
Prompt tags in generative AI prompts are used by the Document Extraction engine to enable specific extraction logic that helps to accurately extract data from complex layouts.
Using prompt tags might be useful in specific scenarios where you do not have to experiment using complex prompts to extract specific information.
From v.37 release, you can select prompt tags when adding prompts to add the tags automatically instead of manually adding the tags. If you had previously added prompt tags manually, the tags would continue to work as expected.
Benefits
Using prompt tags in the generative AI prompts for data extraction provides the following benefits:
- Improved efficiency: You can use prompt tags to easily extract information from separate detectable tables and from specified tables.
- Predefined prompt tags: These tags are specifically introduced to handle data extraction in complex scenarios such as linked tables, signatures, and field relationships.
Prompt tags support matrix
The following table provides the list of prompt tags supported in Document Automation:
Prompt tag | Description | Prompt tag used in | Example prompt | Supported package version |
---|---|---|---|---|
GenAIVisionPlus | Use this tag to indicate to the Document Extraction engine
to use advanced vision-powered generative AI models for better data extraction in scenarios such as table
spanning across multiple pages with headers only on the first page,
tables within a table, and tables without a proper structure. Note: Using this prompt tag might impact the
processing time.
|
|
The following prompt uses advanced vision-powered generative AI models to extract product information from documents containing tables spanning across multiple pages with headers only on the first page.
|
3.37.4 |
GenAIVision | Use this tag to indicate to the Document Extraction engine
to use vision-powered generative AI models
for data extraction. This tag is particularly useful to extract data in complex scenarios such as linked tables, tables spanning multiple pages, nested tables, and merged cells. Note: Vision-powered generative AI models are not supported
for form fields in the unstructured document
type. |
|
|
3.35.14 |
LinkingField | Use this tag to indicate to the Document Extraction
engine that a table field can be used to link separate detectable
tables. Adding this tag ensure that a new column is created in the output file for the linking field. |
Table field |
The following prompt uses vision-powered generative AI models to extract patient names from documents that contain separate tables for each patient containing patient information. In this case, the patient names are extracted in a separate column in the output file.
|
3.35.14 |
TableIdentifier *Table title* | Use this tag to indicate to the Document Extraction engine to differentiate specific tables. The table title is as defined in the document to be extracted and not the table name defined in the learning instance. | Table prompt |
|
3.35.14 |
More resources
To learn more and for examples, search for the Vision Powered Generative AI Data Extraction course in Automation Anywhere University: RPA Training and Certification (A-People login required).