04 — Claims Intake
FNOL
Photos, scans and PDFs converted into structured, source-grounded claim data, validated and handed off to CRM automatically.
The product
Captured from the running systemProduct screen
Product screen
Product screen
How it's built
A first notice of loss arrives as whatever the claimant had to hand — a photo of a form, a scanned PDF, a phone note. FNOL turns that into a structured claim record with each field grounded back to the pixels it was read from.
Validation runs before hand-off: required fields, policy-in-force checks, obvious inconsistencies. A claim that fails validation is queued for a human, not silently accepted.
On success the record is pushed to CRM and the claims workflow with the evidence attached. The adjuster opens a claim that already has its documents indexed.
PythonRustAzure Content UnderstandingAzure OpenAIHubSpot API
On request, not on a schedule
The cluster behind FNOL comes up when someone asks to see it. Ask.