ACORD & loss run data extraction, scored and traceable
RMA reads ACORD 125/127/140, loss runs, and vehicle schedules, and turns them into typed records — every field carrying a confidence score and a link back to its source document.
What RMA pulls from a submission
Not just the easy fields on page one — the whole packet, typed against your schema.
Insured & business
Named insured, DBA, FEIN, mailing and garaging addresses, entity type, years in business, and operations description — pulled from ACORD 125 and the application narrative.
Vehicles & equipment
Power units and trailers from the schedule, with VINs decoded and validated against year/make/model — plus stated values, radius, and use class.
Drivers
Driver roster, license numbers and states, dates of birth, hire dates, and CDL status from supplementals and MVRs.
Coverages & limits
Requested coverages, limits, and deductibles from ACORD 127/140 — auto liability, physical damage, cargo, and general liability.
Loss history
Carrier loss runs itemized row by row: claim dates, status, paid and reserved amounts, and cause — normalized across carrier formats into one table.
Source provenance
Every extracted value links back to the exact document and page region it came from, so a reviewer can verify in one click.
How RMA keeps extraction honest
The risk with AI extraction isn't reading a clean form — it's a wrong value posting silently. Four guards prevent that.
Classify first
Each page is typed before extraction, so an ACORD 140 is read as an ACORD 140 — not guessed at against the wrong schema.
Score every field
Extraction returns a confidence score per field. Anything below your threshold is flagged for a human, instead of silently flowing through.
Validate the value
VINs are decoded, dates and numbers are type-checked, and values are validated against your field schema before they're eligible to post.
Review, then post
A reviewer confirms the flagged fields with the source document side by side. Only approved records reach your policy-admin system.
Document extraction, answered
What document formats can RMA extract from?
Scanned and digital PDFs, images, and spreadsheets — including ACORD 125/127/140 applications, carrier loss runs, schedules of vehicles, MVRs, and supplementals. Mixed and out-of-order packets are classified per page before extraction.
How does RMA handle messy or scanned loss runs?
Loss runs are itemized row by row and normalized across carrier formats into a single structured table — claim date, status, paid, reserved, and cause — even when the source is a low-quality scan.
What is field-level confidence scoring?
Every extracted value carries a confidence score. Fields below your configured threshold are surfaced for human review rather than posted automatically, so accuracy is auditable instead of assumed.
Does RMA decode VINs?
Yes. Vehicle identification numbers are validated and expanded to year, make, and model, and checked for consistency against the schedule — not just transcribed as text.
Watch RMA extract one of your submissions
Bring a real ACORD packet and loss run. We'll run it live and show you the fields, the confidence scores, and the source links.
Book a demo