Limitations
Process-related limitations
This solution can handle a wide range of scenarios, but it is important to be aware of the process limitations in its current version. Understanding these constraints helps ensure a smooth workflow and clarifies when manual intervention in the Review Center is required:
- Multi-PO Invoice Processing
This solution handles invoices with multiple Purchase Order (PO) numbers only if the PO numbers appear on the item line level.
- PO Numbers at the Header Level
If multiple PO numbers are listed in the header section of the invoice, the system processes only the first PO number it identifies. The invoice matches against that single PO, and any discrepancies or remaining balances require manual handling.
- Unsupported invoice types
The solution currently does not support Financial Agreement Invoices. However, it can process Purchase Order (PO) Invoices, Expense Invoices, and Credit Notes.
- U/M conversions
The system does not automatically convert Units of Measure (U/M). However, you can enable U/M validation by configuring the enableUOMValidation argument in RPA. When this argument is set to True, the U/M extracted from the invoice is validated against the corresponding Purchase Order (PO) line. If a mismatch is detected, the invoice is routed to the Review Center for manual confirmation. For example, if the invoice uses case while the PO uses each.
- Discounts
The automated process does not apply or calculate discounts specified on the invoice, whether at the header or line level. The system ignores any discount information and processes the invoice for its gross amount. It does not send an exception to the Review Center for discounts; any potential early payment discounts remain unprocessed unless handled manually outside this workflow.
- Complex matching scenarios
When the system cannot determine a unique match for an invoice line, this logic is applied:
- Lines with Invoiced status are excluded.
- If duplicate lines have matching Item Code, Supplier Item Number, PO Line Item Name, PO Line Description, price, and quantity, the system prioritizes the line with the lowest sequence number for matching.
- Each invoice line can match only one purchase order (PO) line. The system does not support splitting a single invoice line across multiple PO lines. If necessary, you can manually adjust or replace the match in the Review Center.
- Language support:
- While the underlying IDP with GenAI has broad language capabilities, this solution is fully tested only for documents in English.
- Error messages and labels in the Review Center UI are available in English only.
Document capture limitations
Documents must meet these requirements to be processed in the IDP system:
- Supports only the PDF, PNG, JPEG, and JPG file formats as an attachment to email or files placed in a local folder. These file formats compressed in a ZIP file are also supported. XML-based or Excel-based formats are not supported.
- Processes files up to 20MB; however, the allowed file size might vary depending on the Infor OS tier level.
- Processes PDFs with up to 99 pages.
- Requires PDF quality of more than 300 DPI.
This solution uses Infor Document Processor (IDP) with a generic preconfigured GenAI prompt to extract data from invoices. This approach is designed for high accuracy on standard invoice layouts but has a key limitation regarding customization.
The generic nature of the prompt means it optimizes for standard fields such as Invoice Number, PO Number, and Total Amount but cannot extract non-standard or custom fields that are unique to a specific supplier's template.
Example:
If a standard invoice has a field clearly labeled Invoice Number: INV-12345, the generic prompt can reliably extract this value.
However, if a supplier’s invoice includes a unique, unlabeled reference code that your business uses for internal routing but is not part of a standard field such as Service ID: ABC-987, the system does not capture the ABC-987 value.
As a result, invoices that rely on unique, non-standard fields for processing end up in the Review Center for manual data entry of those specific fields.