Extracted fields are currently returned as raw JSON and displayed without any visual connection to the source document. Users have limited insight into where values originated, how confident the model was, or which fields were generated by AI.
Idea
Explore enhancements to the document review experience that improve transparency and support human validation of extracted results.
Potential enhancements include:
- Display extracted fields directly on the source document using bounding boxes, colour-coded by confidence level, with synchronized navigation between the document viewer and the Fiori form. This would require persisting the bounding box coordinates returned by Document AI.
- Visually distinguish AI-extracted values in the Fiori form (for example, with a badge or icon) so users can easily identify which fields were populated by the model.
- Introduce a human-in-the-loop review workflow in which users can verify, correct, and confirm extracted values before they are finalized. The confirmed results could then be submitted back to Document AI as ground-truth feedback to enable Instant Learning and improve model accuracy over time.
Extracted fields are currently returned as raw JSON and displayed without any visual connection to the source document. Users have limited insight into where values originated, how confident the model was, or which fields were generated by AI.
Idea
Explore enhancements to the document review experience that improve transparency and support human validation of extracted results.
Potential enhancements include: