Document processing and recovery

Resolve partial batches, unreadable content, stale revisions, and uncertain outcomes.

Start by identifying the stage that failed. File intake, document processing, extraction, table review, and acceptance are distinct operations. Repeating the first operation is rarely the right response to every later failure.

Accepted but not ready

Keep the file's intake acknowledgement and inspect its processing feedback. A completed workflow can still be awaiting result rows, and one successful file does not establish that every file in a batch was indexed. Open the authoritative document list when it becomes available.

If processing observation becomes unavailable, the file can remain saved while its progress cannot be confirmed. Check its current result before uploading another copy. Closing the browser does not necessarily stop background initial-draft preparation.

Poor text or unresolved values

Compare extracted text with the original pages. OCR can misread numbers, punctuation, layout, or signatures. A supported file format does not guarantee perfect recognition. Select the appropriate profile, narrow the source set, and review the schema if the proposed fields do not match your purpose.

A missing or ambiguous value should remain unresolved. Do not fill it by assuming a nearby date, amount, or party belongs to the field.

Stale revisions

A conflict means the document, table, or draft changed after your view was loaded. Reload the current state and compare it before applying another change. A stale approval must not be reused for a replacement revision.

Lost responses

An uncertain request may already have committed. Recover the saved draft, document revision, or deletion receipt first. Use the existing retry or reload path for that action instead of starting another paid generation or deletion intent.

If evidence verification fails, do not substitute the current original as proof of an earlier value. Start a new explicit review if the source basis must change.