From the first "hello" to an issued policy, fully autonomous, audit-ready, human when it matters.

FAMI takes a customer from conversation to policy or payout with almost no manual work in between. Four agents, one pipeline, one live dashboard.

Request a demo See the 90-second demo

What FAMI is

FAMI runs the whole motor insurance lifecycle, conversation, evidence, underwriting, issuance, as one traceable pipeline. When something needs a human, it routes there by design.

GoDo talks to the customer. ENOL reads the vehicle from photos. Zentis underwrites under guardrails. The FAMI core ties all three into one event-tracked flow.

Quote-to-issue, before FAMI

A call gets routed through an IVR, photos arrive by email and get re-keyed by hand, and the file sits in an underwriting queue for days. No single record shows who touched what, or when.

Industry Friction

The pain points we hear most

From carriers, MGAs, and InsurTechs trying to move faster without losing the paper trail.

Intake is the bottleneck

Phone trees and email attachments cost days before a file is even underwriting-ready.

Damage reads depend on who's looking

Two adjusters read the same photos two different ways.

Underwriting is a black box or a bottleneck

Rules alone can't judge; manual review alone can't scale.

Channels don't talk to each other

Start on WhatsApp, finish by phone, repeat everything.

The Solution Architecture

How FAMI solves it

One pipeline, four agents, every step logged as it happens.

One AI, every channel

GoDo carries the same conversation across voice, webchat, and WhatsApp.

Documents in strict order

License, registration, prior insurance, collected step by step, no skipping.

The vehicle reads itself

ENOL checks damage, identity, and odometer from five photos.

Underwriting runs guardrailed

Zentis runs the checks and fails closed on anything uncertain.

Every step is a tracked state

The orchestration core streams each transition to the live dashboard.

Stakeholder Value

Built for the person who owns the funnel

Different roles, different stakes, all covered by the same testing layer underneath.

Head of Digital / Innovation

The speed question

Days of hand-offs become one conversational session.

VP Underwriting

The consistency question

Every file gets the same evidence and the same checks.

Head of Claims

The evidence question

Damage read, identity check, and transcript, one traceable record.

Foundations

The four pillars

Purpose-built, deterministic, and auditable, working as one pipeline.

01 · Conversation

GoDo

Voice, webchat, and WhatsApp intake for policy and claims.

02 · Evidence

ENOL

Damage and identity reads from five structured photos.

03 · Underwriting

Zentis

SAFE-guarded underwriting, visual and repeatable.

04 · Orchestration

FAMI Core

Event bus and state machine tying the three together.

Core Capabilities

Features

Three channels, one assistant

Voice, webchat, and WhatsApp share one conversation state.

Auto-extracted identity

License uploads read automatically, no manual typing.

OTP-verified claims

Identity confirmed before any incident detail is captured.

Structured vehicle evidence

Five angles, stored in S3 behind short-lived URLs.

Vision-LLM damage reports

Provider agnostic: Qwen, Gemini, Claude, or OpenAI.

SAFE guardrails

PII masking, no invented facts, fails closed on error.

Full phase replay

Every phase stores its artifacts for later review.

Real-time dashboard

Live WebSocket view of every event, policy and claim alike.

Benchmark Comparison

FAMI against a manual funnel

Dimension	Manual quote-to-issue	With FAMI
Intake	IVR hand-offs, email, re-keying	One conversation, staged documents
Damage read	Damage read varies by adjuster	Consistent vision-LLM damage read
Underwriting	Underwriting queue, checks when time allows	Underwriting runs on arrival, guardrailed
Record	Record scattered across email and calls	One event-tracked, replayable record
Turnaround	Days to an issued policy	A session, tracked live on the dashboard

Infrastructure

Architecture and security

A RabbitMQ event bus and async workers move policies through a Postgres-backed state machine. Documents and reports persist in S3 behind short-lived URLs. Conversations run over WebSocket and Twilio with Redis state. Vision analysis stays provider-agnostic, and OpenTelemetry traces the whole pipeline for replay.

Tech stack

FrontendsCLIENT

React & TypeScript

React 18, Vite, Tailwind, React Flow.

ConversationsREALTIME

Voice, chat, WhatsApp

WebSocket, Twilio, Redis sessions.

OrchestrationSTATE MACHINE

Event-driven core

RabbitMQ, async workers, Postgres.

Vision & storageMULTIMODAL

Evidence layer

Vision-LLM providers, AWS S3.

See it in a real workflow

Chat to issued policy in one session

"I want to file a vehicle policy"

GoDo takes phone and email, reserves the policy.

T+0s

License upload

Name auto-extracted, confirmed back to the customer.

T+15s

Registration and five photos

Collected in sequence.

T+35s

ENOL reads the vehicle

Damage, identity, and odometer, returned as a structured report.

T+50s

Coverage picker

Plan options render inline, customer picks one.

T+65s

Zentis underwrites

Checks run under guardrails, audit events shown live.

T+80s

Policy issued

BIND COMPLETED

Payment link sent, full timeline on the dashboard.

T+90s

Quantifiable Impact

What this could mean

Illustrative. The real number depends on line of business and scope.

Days → One sessionFirst contact to issued policy

3 channels → 1 agentVoice, webchat, WhatsApp, shared state

LiveAudit trail, streamed as it happens

Who it's for

Insurance carriers & MGAsInsurTech & mobility startupsTeams that need an auditable workflow, not a black box

Engagement

Where to start

Bring one line of business. We'll file a real policy live, then pull the audit trail so you can see exactly what happened.

PILOT VERIFICATION

Want a live walkthrough? We'll file a real policy in 2 minutes, then pull the audit trail.

Request a demo