03
Claim vision

Photo to first notice in seventy seconds.

A customer texts a photo of bumper damage. Seventy-four seconds later, the FNOL is drafted, the plate is extracted, the policy is matched, and the broker is one click from filing.

74s
median photo → FNOL
from SMS hit to draft
412
claims auto-filed today
across the network
1.4s
vision latency
per-image, p50
94%
extraction confidence
on damage classification
Live preview

Photo to first notice in seventy seconds.

bob.so/app · claims · CLM-2026-1217
live
img_0412.jpg
PlateABC123
Damagerear bumper
Severitymoderate
Odometer84,210 km
FNOLdrafted · ready
Multimodal

Plates. Damage. Odometer. Signs of injury.

Bob's vision model is trained on insurance-specific imagery. It pulls the same facts a human assessor would and writes them straight to the policy's fact-store entry.

Capture

Customers don't open an app.

Audit trail

Every frame, hash-anchored.

The raw photo, the extracted facts, the prompts, the FNOL — all four hash-chained into the audit ledger. Open the claim in 2030 and the original frame is right there.

The next twenty insurance brokerages will not be built on spreadsheets.

We’re onboarding ten design-partner brokerages this quarter. If your book is between 1,000 and 12,000 policies, we should talk.

b/
bob
AI-first insurance CRM. Bob runs the operational layer — voice, vision, documents, renewals, claims, compliance — so brokers can do the broking.
© 2026 Bob — AI-first insurance CRMThinkspace · 30 Pollen Street · Grey Lynn · Auckland