Configure Enhanced Invoice Learning

The enhanced Learning UI provides one workspace for reviewing and improving invoice-document mappings for unstructured documents.

The enhanced Learning UI provides one workspace for reviewing and improving invoice-document mappings for unstructured documents.

Payables Agent can process invoices using AI-based recognition before training is performed. Training is typically completed once per invoice layout, with priority given to high-volume suppliers, recurring layouts, and suppliers that generate frequent recognition issues.

Use Adaptive Learning for Descriptive and Global Descriptive Flexfields in Edit Invoice

From release 26C onward, adaptive learning in Edit Invoice supports additional invoice attributes at the header and line levels for image invoices. During training, all labeled attributes are extracted from source documents and can be mapped to invoice attributes, including descriptive flexfields and global descriptive flexfields, which weren't previously supported. You can train customer-specific business attributes captured through Descriptive flexfields and country-specific localization and compliance attributes captured through Global descriptive flexfields. Your training is saved as learning data and automatically applied to subsequent invoice documents. This reduces repetitive manual intervention, improves attribute accuracy over time, and streamlines downstream invoice processing. Where context is required, users will be able to provide context-sensitive instructions that help the application resolve flexfield segments and regional attributes.

For PDF and image invoices, the Learning UI provides an annotation-based experience. Users review the source document beside the invoice preview and identify missing or incorrect values directly on the document.

Saved corrections are associated with the document fingerprint. When applicable, the learning is reused for future documents and can be propagated to other invoices with the same fingerprint, reducing repeated corrections.

To configure and validate learning:

1. Select the representative unstructured documents.

2. Open the document in the Learning UI.

3. Review the source attributes and mapped invoice attributes.

4. For image or PDF invoices, annotate missing or incorrect values where required.

5. Save the learning.

6. Reprocess an applicable document or submit another document with the same fingerprint.

7. Confirm that the corrected mapping or transformation is reused.