Your adjusters spend 14 hours per week searching for documents. AI classification fixes that.
Regure's document processing engine ingests every document from every source — email, upload, fax, API — classifies it with 99%+ accuracy, extracts key fields, and attaches it to the correct claim file. No manual sorting. No lost documents. No data entry.
Insurance runs on documents — and those documents are everywhere
A single auto claim generates 8-15 documents: the FNOL form, police report, photos of damage, repair estimate, medical records (if injuries), claimant statement, witness statements, the policy declaration page, prior loss history, and settlement documentation. A complex commercial property claim can generate 50+ documents across months of investigation.
These documents arrive through every channel imaginable. The police report comes as an email attachment. The repair estimate is faxed from the body shop. Photos are texted to the adjuster's personal phone. Medical records arrive by mail and get scanned at the branch office. The ACORD certificate is submitted through the broker portal. The claimant statement is recorded over the phone and transcribed.
Without automated processing, adjusters become document clerks — spending 40% of their workday searching for, sorting, and manually entering data from documents instead of assessing claims and making decisions. This is the single largest source of claims cycle time inefficiency.
Regure's AI document processing engine eliminates manual document handling entirely. Every document is ingested, classified, data-extracted, and routed to the correct claim file — in seconds, not hours.
AI document classification trained on insurance document types across all lines of business
Regure's classification engine doesn't use keyword matching or filename patterns — it analyzes document structure, content layout, and semantic meaning to classify documents with the accuracy insurance operations require.
Generic document classification tools fail on insurance documents because insurance has unique document types that don't exist in other industries. A bordereaux is not a spreadsheet. An ACORD 25 certificate is not a generic PDF form. A loss run report is not a financial statement. Insurance document classification requires models trained specifically on insurance documents.
Regure's classification engine handles 20+ insurance-specific document types out of the box:
- FNOL forms and loss notices — ACORD and carrier-specific formats
- Police reports and incident reports — structured and unstructured formats from different jurisdictions
- Medical records — treatment notes, surgical authorizations, discharge summaries, billing statements
- Repair and replacement estimates — body shop estimates, contractor bids, equipment replacement quotes
- ACORD forms — certificates of insurance (25/28), applications (125/126/130/140), loss notices
- Policy documents — declaration pages, endorsements, policy jackets, binders
- Legal documents — subrogation letters, settlement releases, demand letters, court filings
- Financial documents — payment receipts, invoices, salvage valuations, reserve worksheets
- Photos and images — vehicle damage, property damage, injury documentation, scene photos
- Correspondence — claimant communications, adjuster notes, broker transmittals, regulatory notices
Classification confidence scores accompany every result. Documents classified with high confidence (above 95%) route automatically. Documents with lower confidence queue for human review — maintaining accuracy without creating bottlenecks.
Key field extraction that eliminates manual data entry from claims processing
Classification tells you what the document is. Extraction tells you what's in it. Regure extracts structured data from unstructured documents — claimant names, policy numbers, loss dates, amounts, and dozens of other fields — and maps them to your claims system.
ACORD Form Extraction
ACORD forms are the backbone of US commercial insurance — but data still gets manually re-keyed from ACORD PDFs into management systems. Regure extracts data from ACORD 25, 28, 125, 126, 130, and 140 forms with field-level accuracy: insured name, policy number, effective dates, coverage limits, deductibles, and named additional insureds.
For US MGAs and brokers, ACORD extraction eliminates the manual data entry that adds 15-30 minutes per submission — across hundreds of submissions per month.
Medical Record Extraction
Health and casualty claims require data from medical records: treatment dates, diagnosis codes (ICD-10), procedure codes (CPT), treating physician information, and treatment costs. Regure extracts structured data from medical treatment notes, surgical authorizations, discharge summaries, and billing statements.
For carriers processing health claims, medical record extraction reduces the manual review time that bottlenecks claim assessment — getting structured data into the claims system while the adjuster focuses on medical necessity evaluation.
OCR for Scanned & Handwritten Documents
Not every document arrives as a clean digital PDF. Police reports from small departments arrive as scanned carbon copies. Claimant statements are sometimes handwritten. Older policy documents were never digitized. Regure's OCR engine handles degraded scans, handwritten text, and multi-language documents with high accuracy.
OCR output feeds directly into the classification and extraction pipeline — a scanned police report is treated identically to a digital PDF, with the same field extraction and claim linking.
Photo & Image Analysis
Damage photos are critical evidence in property and auto claims. Regure analyzes photos to extract metadata (geolocation, timestamp, device information), detect damage severity indicators, and identify photo manipulation. For auto claims, vehicle identification from photos cross-references against claim VIN data.
Photo analysis doesn't replace adjuster assessment — it provides structured metadata that helps adjusters evaluate claims faster and flags inconsistencies (photos taken at a different location than the reported loss, for example).
Duplicate detection, missing document alerts, and claim file completeness
Processing documents is only half the challenge. Knowing what's missing, what's duplicated, and whether a claim file is complete for assessment — that's intelligence. Regure monitors claim file completeness and acts on gaps automatically.
Duplicate Detection
The same document submitted multiple times — by the claimant, the broker, and the adjuster's field notes — creates confusion about which version is authoritative. Regure detects duplicates using content hashing (exact matches) and similarity analysis (near-duplicates like rescanned versions of the same document).
Duplicates are flagged but not deleted — the system marks the most recent or highest-quality version as primary while retaining all copies for audit trail completeness.
Missing Document Alerts
Every claim type has required documents. An auto collision claim needs a police report, damage photos, and repair estimate. A health claim needs a medical authorization and treatment records. Regure knows what's required per claim type and jurisdiction — and auto-generates document requests when gaps are detected.
Missing document alerts go to the responsible party (claimant, broker, field adjuster, or medical provider) with specific requests and deadlines — before the assigned adjuster has to chase them manually.
Claim File Completeness Score
Before an adjuster begins assessment, they need to know whether the claim file is complete. Regure provides a completeness score per claim — showing which required documents are present, which are missing, and which are pending receipt. Adjusters prioritize claims with complete files, avoiding the stop-start workflow of beginning assessment only to discover a missing document.
Completeness scoring is configurable per line of business: auto claims require different documents than property, health, or workers comp.
Ingest documents from any source — email, upload, API, fax, or mobile app
Documents arrive through every channel. Regure monitors all of them simultaneously, processing documents identically regardless of how they arrive.
Email Ingestion
Configure monitored email addresses (claims@, FNOL@, documents@) where incoming emails and attachments are automatically processed. Regure extracts attachments, classifies each document, identifies the related claim from email content (policy numbers, claim numbers, claimant names), and attaches to the correct file. The email itself is logged as correspondence.
For brokers who submit via email, this eliminates the manual download-classify-attach process that operations staff perform hundreds of times per day.
API & System Integration
Broker management systems (Applied Epic, Vertafore AMS360), carrier portals, and aggregator platforms submit documents via Regure's REST API. API submissions include structured metadata — making classification instant and field extraction unnecessary for pre-structured data.
Bi-directional APIs with core claims systems (Guidewire, Duck Creek, Sapiens) ensure document data flows into your claims system without manual re-entry. Claim status updates flow back to Regure for portal display and reporting.
Mobile App Capture
Field adjusters photograph damage, record audio notes, and capture signed statements directly from the Regure mobile app. Photos are geotagged and timestamped automatically. Audio recordings are transcribed and classified. Signed statements capture digital signatures with witness verification.
All mobile captures sync to the claim file immediately when connectivity is available — and queue locally for sync when field adjusters are in areas without cellular coverage.
Fax & Scan Processing
Medical providers, legal offices, and some government agencies still send documents by fax. Regure's fax digitization converts incoming faxes to searchable PDFs, applies OCR, classifies the document, and routes it to the correct claim file. Scanned documents from branch office scanning stations are processed identically.
Fax and scan processing ensures no document falls through the cracks regardless of how legacy-dependent the sender's technology stack is.
What operations teams ask about document processing
How does AI classification achieve 99%+ accuracy?
Regure's classification engine is trained on millions of insurance-specific documents across all lines of business. It analyzes document structure, content layout, and semantic meaning — not just keywords or filenames. Confidence scoring routes uncertain documents to human review, maintaining accuracy while preventing bottlenecks. The model improves continuously as corrections feed back into training.
Can Regure handle non-English documents?
Yes. OCR and classification support English, Spanish, French, German, Arabic, and other languages common in multi-market insurance operations. For Middle East operations, Arabic/English bilingual document handling processes documents in both languages with the same classification accuracy.
What happens when classification confidence is low?
Documents below 95% confidence queue for human confirmation — a one-click classification that takes 5 seconds per document. Documents below 80% route to manual classification. Over time, human corrections improve the model's accuracy on edge cases. Most operations see 95%+ of documents classified fully automatically within the first month.
Does field extraction work on handwritten documents?
Yes, with appropriate expectations. Regure's OCR handles handwritten text with accuracy that varies based on legibility. Clear handwriting achieves 90%+ character accuracy. Degraded handwriting flags for human review. For insurance operations, the most common handwritten documents are claimant statements and older claim notes — both of which Regure processes effectively.
How does Regure handle document versioning?
When multiple versions of a document exist (revised repair estimates, updated medical records), Regure maintains the complete version history with the most recent version marked as primary. Previous versions remain accessible for audit purposes. Duplicate detection distinguishes between true duplicates (same document submitted twice) and version updates (revised documents that supersede earlier versions).
What's the processing speed?
Standard documents (PDF, DOCX) classify in under 3 seconds. OCR-dependent documents (scans, faxes, photos) process in 5-15 seconds depending on page count and image quality. Bulk uploads process in parallel — 100 documents complete in under 2 minutes. Enterprise tier includes priority processing for time-sensitive documents.
See Regure classify your actual insurance documents
Book a 20-minute demo. Bring your real documents — police reports, ACORD forms, medical records — and watch Regure classify, extract, and route them in seconds.