Pre-process
Cut 200-page scans into clean documents
Boundary detection splits mixed PDFs into single-doc jobs with cited split points. Analysts stop sorting pages by hand.
Split once, extract clean
01
Up to 200pp / 100MB
Loan files, AP inboxes and claim packs in one upload.
02
Cited boundaries
Every cut carries page + reason — rejoin or correct from the viewer.
03
Mixed scans welcome
Rotated phone photos and interleaved pages sorted automatically.
The split pipeline, stage by stage
Follow a 60-page file from upload to three clean jobs.
Step 3 of 4: Route, due Instant.
Layout and content shifts mark candidate boundaries across all pages.
FaFieldry automation owns this step
Boundaries confirmed with confidence; low-confidence cuts flagged for review.
FaFieldry automation owns this step
Each piece flows to Classify, then to its schema — one pipeline per doc.
FaFieldry automation owns this step
Drag a boundary in the viewer; downstream jobs re-run on the fixed ranges.
YtYour team owns this step
Detect
Layout and content shifts mark candidate boundaries across all pages.
Cut
Boundaries confirmed with confidence; low-confidence cuts flagged for review.
Route
Each piece flows to Classify, then to its schema — one pipeline per doc.
Correct
Drag a boundary in the viewer; downstream jobs re-run on the fixed ranges.
From scan pile to clean jobs
01
Upload the pile
One PDF with invoices, statements and IDs mixed together.
02
Get the cuts
Split returns ranges with reasons — pp.1–12 invoice, pp.13–40 statement.
03
Extract per doc
Each range extracts against its own schema with full evidence.
Related use cases
Layout breaks, header/footer changes and content-type shifts combine into candidate cuts, each with confidence and reason.
Correct it in the viewer — drag the boundary and downstream jobs re-run. The correction is logged.
200 pages and 100MB per file for split. Larger files go through batch pipelines up to 5GB.
Split itself carries no surcharge — extraction meters per resulting page (4/page, +2 with tables).
Cut scans into documents
Cited boundaries, correctable.
Three loan docs in one scan become three clean jobs — one drag fixes a bad cut.