Insurance claim document processing
Tavnit processes a claim and the evidence supporting it as one set: the claim form, repair estimates, invoices and reports. Each document type is routed to its own flow, and the results assemble into a single record a handler can assess rather than a folder they have to read.
Why this is painful
A claim is never one document. It is a form plus whatever the claimant sent — estimates, receipts, a police report, photographs — arriving over days, in no order, by email.
Handlers spend most of their time assembling rather than assessing: opening attachments, finding the amounts, checking that the figure claimed matches the figure on the estimate.
What to extract
| Field | Why it needs care |
|---|---|
| Claim and policy number | The keys everything in the set joins on, and often the only thing consistent across documents. |
| Claimant and incident details | Date, location and description of loss, usually spread between the form and a narrative. |
| Amount claimed | What the form says. Worth extracting separately from what the evidence supports. |
| Supporting amounts | Estimate and invoice totals — the figures the claimed amount should reconcile to. |
| Third-party details | Other parties, insurers and reference numbers, needed for recovery. |
| Document type per attachment | Knowing what each file is determines which flow reads it. |
What makes insurance claims hard
Evidence arrives late and out of order
Assessment cannot start until the set is complete, and the set completes over days. Collections classify each attachment as it lands so the record builds incrementally, rather than someone re-reading the whole folder each time something new arrives.
The claimed amount and the evidenced amount are different fields
Extracting one number per claim loses the check that matters. Capture what was claimed and what the estimates and invoices actually total, and the discrepancy becomes a value you can filter on rather than something a handler has to notice.
Claim files are personal, and often medical
Claims routinely contain health information, identity documents and financial detail. Private Buckets with per-user access grants keep a claim visible to the handlers on it rather than to everyone in the organisation.
How the pipeline handles it
- Collections — Classifies each attachment — form, estimate, invoice, report — and routes it to the flow built for it.
- Email Integration — Claims arrive by email; forwarding to a Collection address processes every attachment as it lands.
- Cleaners — Normalises dates and currencies, and computes the gap between claimed and evidenced amounts.
- Human in the Loop — Handlers approve or reject with a permanent record of the decision and who made it.
Common questions
- Can it handle a claim with many different attachments?
- Yes. A Collection classifies each attachment and routes it to the right flow, so a claim form, an estimate and an invoice are each read by the flow built for that document type.
- Can it flag claims where the numbers do not agree?
- Extract the claimed amount and the supporting totals as separate fields, then use a Cleaner to compute the difference. Claims that fail the check can be routed to review automatically.
- How is sensitive claim data protected?
- Buckets can be private rather than organisation-visible, with per-user Viewer or Editor grants, so a claim is visible only to the handlers working it.
Other document types
Try it on one document
Build a flow, send a real insurance claim through it, and see what comes back. Free credits to start.
Start free