Invoice capture, file filters applied at the file name level are ineffective because the AI also analyzes the document body content.

Based on its name, the expectation is that exclusion words would be validated only against the file name. However, we’ve observed that the validation process is also scanning the document’s content. If an exclusion word is found within the body of the document—even if it’s not present in the file name—the document is rejected.

This behavior contradicts the intended functionality and can lead to unnecessary rejections. It would be beneficial to either restrict the validation strictly to the file name or provide configuration options to control whether content should be scanned.