Capability · Document Intelligence
Complex documents,
read properly the first time.
Handwritten statements, 200 pages, six agreements in one PDF (three irrelevant). Document Intelligence reads scanned files OCR misses, extracts data with 99%+ accuracy, and links each extraction to its page for easy review.
What firms running Claim Inbox are seeing
- Documents processed
- 5m+
- Longest medical record
- 2,200pp
- Faster per document
- 90%
What it reads
The documents that break ordinary OCR.
The files a group action actually runs on, not clean, templated PDFs, but the real thing.
DSARs & finance agreements
Multi‑agreement bundles from lenders: several agreements in one scanned PDF, each pulled out and matched to a case.Medical & GP records
GP records from 94 to 2,200+ pages, handwritten drug lists and all, read for the details that matter to the claim.MOD & service records
Service and medical records, often decades old and handwritten, read for the qualifying facts, with the page noted.Audiograms & reports
Audiograms, surveyor reports, third‑party letters: classified, extracted and filed against the right case.Handwritten & scanned
Image‑based and handwritten pages that traditional OCR misses entirely: read, not skipped.Two hundred pages
Long bundles where the relevant fact is on page 137: found, extracted and anchored so a person can verify it fast.
How does it work
Read, extract, anchor;
then a person checks it.
Ingest
Every document arrives from the inbox, a portal or the case file, whatever format, whatever length.
Read
Scanned, image‑based and handwritten pages are read where ordinary OCR gives up.
Extract
The fields that matter are pulled into structured data at 99%+ accuracy in production.
Anchor & check
Every extraction is linked to its source page, so a fee earner verifies it in one click before it’s used.
- Read: A 200‑page scanned bundle, ingested whole.
- Check: Each field anchored to its page, confirmed in one click.
Across claim types
One capability, every kind of document.
PCP & motor finance
Agreements extracted from scanned DSAR bundles, several to a file, each pulled out and matched to the right case on registration and agreement number.
NIHL & industrial disease
GP and MOD records read for ototoxic medication, audiograms and qualifying facts: a review that takes a paralegal one to two hours, done in minutes.
Personal injury & housing
Medical reports, surveyor bundles and third‑party letters classified and extracted: the evidence gathered before anyone opens the file.
The rest of the platform
Six capabilities, one operation.
Every capability is a block in the Saber Workflow Builder; here’s the whole platform, and where you are in it.
Document Intelligence
You are hereReads DSARs, medical and MOD records (handwritten, scanned, hundreds of pages) and extracts every field, anchored to the source.
Correspondence Processing
Inbound mail classified against the right case, the request actioned and the reply drafted: the queue that clears itself.
Client Communications
Updates, acknowledgements and query answers drafted and sent, on the file, in your voice, with a person approving each.
Claim Intelligence
Eligibility and merit assessed from FOS outcomes and your own data; the right claims move, the weak ones are caught early.
Paralegal Performance
Every handler’s throughput, live: who’s flying, who’s stuck, where to coach.
Book Intelligence
Live portfolio visibility: tranche performance, where the book is stuck, what to chase.
Automate your book
with Saber.
Send us one workflow you run today: a DSAR, a batch of lender mail, a round of client updates. By the end of the week we’ll show it read, actioned and ready for a paralegal to approve, on your data. No fee, no commitment.
Extraction you can defend
Every figure is anchored to its source page and a person approves it; nothing enters the case unchecked. Full audit trail, ISO 27001 and ISO 42001 certified, UK‑only data residency, never used to train our models.
