Use case
Claim packs with forensics signals before payout
FNOL forms, estimates and invoices parsed to schema — edited files and mismatched totals flagged first.
The problem
A claim is a folder, not a file — and the payout decision rests on all of it holding together.
- 01Pack chaos. FNOL forms, contractor estimates, prior invoices and ID copies land unsorted in one folder.
- 02Inflated estimates. Quotes and invoices disagree with each other — and sometimes with the policy itself.
- 03Invisible edits. Altered line items and resized scans pass the naked eye at volume.
- 04Payout audit trail. Every settled claim must show its working years later. Screenshots don't qualify.
The solution
Fieldry structures the whole pack, checks it against itself, and hands adjusters cited values plus tamper signals.
- 01Whole-pack parsing. Forms, estimates and invoices each extracted to schema, linked to the same claim.
- 02Forensics pre-screen. File, image and consistency checks run before payout — edits and duplicates flagged.
- 03Consistency checks. Totals versus rows, dates versus policy periods — contradictions surface both sides.
- 04Adjuster evidence. Each approved value links to page and bbox; the payout file defends itself.
Extraction checks
Search the claims checks.
See what each check returns.
Every check Fieldry runs on a claim pack.
Whole-pack parsing
Forms, estimates and invoices each extracted to schema, linked to the same claim.
Forensics pre-screen
File, image and consistency checks run before payout — edits and duplicates flagged.
Consistency checks
Totals versus rows, dates versus policy periods — contradictions surface both sides.
Command preview
Linked records
Each document in the pack extracted to schema and linked to the same claim file.
Anatomy of a claim file
A storm claim arrives: FNOL form, two contractor estimates $4,000 apart, a prior invoice and the policy schedule.
Fieldry parses all four, flags the estimate gap and a duplicated prior invoice — the adjuster settles from evidence, not email threads.
- InputFNOL + estimates + invoice + policy
- OutputCited values + tamper flags
- ReviewAdjuster decides from evidence
From claim folder to defensible payout
Adjuster settles from cited values; the file carries its own defense.
AnalystForensics and consistency checks flag the gap and the duplicate prior.
FieldryDocuments parsed to schema and linked; estimates compared line by line.
FieldryClaim pack lands unsorted — forms, estimates, priors in mixed quality.
Document
4 milestones, newest first.
Key benefits
Pay clean claims fast, hold the suspicious ones — with proof either way.
Straight-through clean
High-confidence packs settle untouched; adjusters work the flagged queue.
Ring signals
Duplicates across claims surface before the second payout, not after the tenth.
Cheap bulk
Seasonal piles drain through batch at half price; urgent claims stay sync.
Regulator-ready
Every decision value carries its source for complaints and audits.
Hold the next bad payout
Run a real claim pack on free credits and see which values the forensics layer flags.
Get startedRelated use cases
Related industries
File anomalies (edited PDFs), image anomalies (resized or copy-pasted regions) and consistency breaks (totals vs rows). Signals, not verdicts — adjusters decide.
No — Fieldry produces signals and cited values. Judgment stays with your adjusters.
Yes — repeat documents and matching line patterns across claim files flag before the second payout.
REST for real-time intake, webhooks for async bulk, n8n for no-code. Results carry the claim ID you passed in.
Parse your first document today
One API for extraction, evidence and automation.