If you have attached or uploaded a file to a file attachment field that has been configured to run AI analysis, and you are not seeing results in the output field, read on for some common troubleshooting steps.
- Confirm you have saved the file attachment. It is the act of clicking Save that sends the file off for analysis.
- Confirm the file has been OCRed. All files need their text extracted before they can be sent for analysis, and this happens in the OCR step. You can see if a file has been OCRed by going to the Properties section for the document. If the file is marked as queued for OCR or in progress, you will need to wait until it has been completed for analysis to begin.
If the file has OCRed and AI analysis is still in progress:
- AI analysis is not instant. It will typically take 4-5 seconds per page to process, and could take 30 minutes or more for a longer document. You may need to move on to other tasks and return to the field at a later time.
- Check the status of the job. You can do this by going to Report Builder and running the AI Field Job List report.
- Check that your AIFields are configured properly.
- If you are working in a static section, the file attachment field and the multi-line text field do not have to be in the same location. If you are working in a collection section, however, the two fields must be in the same section.
- If you are a FV admin and have access to the Customs Editor, doublecheck your settings in the Customs Editor. Make sure the AIField is linked to the correct file attachment field and has the appropriate analysis type.
If OCR has failed, but the document has been indexed with a high degree of confidence (shown as a percentage in the file properties), AI analysis may still be successful. If, however, both OCR and indexing have failed, proceed to the subsequent section on OCR Troubleshooting.
Note: If you cannot see OCR details in the Properties panel, it's possible OCR has not been turned on for your Org. Please reach out to support to confirm Docs+ is enabled for your firm.
OCR can sometimes fail due to various reasons, including:
- Poor image quality or low resolution
- Complex fonts and handwriting
- Noisy background
- Skewed or rotated text
- Inconsistent formatting
- Incorrect or suboptimal scanning settings
- Document size
OCR tends to work best with well-structured and consistent documents that are ~2000 pages or less. If the document lacks proper formatting, it’s possible OCR could struggle to differentiate between elements such as headers, footers, and body text, especially within forms. If OCR has failed for a given document, try the following:
- Occasionally OCR will fail but the document will be sent to our backup method. In this scenario, there will be a confidence percentage listed within the document properties. If there is a high percentage of confidence, it is likely AI analysis will still occur.
- If it’s a larger document (exact page count can vary based on overall document complexity), it may be useful to divide the document into smaller segments and attempt to upload those separately into Filevine.
- Download the document, and then open it print to a PDF (print the document, and select Save as PDF as the destination). Once the document is saved to your device as a PDF, you can reupload the document as a new doc or version.
- When you print a document to PDF, the content is converted into a digital format with a standard resolution and image quality. This can result in a clearer and more legible document, which can enhance OCR accuracy compared to a low-quality scanned image.
- Printing to PDF creates a standardized digital document with consistent formatting, making it easier for OCR software to interpret the content correctly.
- It's important to note that if the original document has complex elements like handwritten text, complex layouts, or poor image quality, printing it to PDF may not completely resolve OCR issues.
If the issue persists after these troubleshooting steps, reach out to Filevine support.