FNOL Automation · Dutch Non-Life Insurance

It's Monday morning.
Your FNOL team just started their shift.

300 calls are coming in. Each one touches 4 disconnected systems. Each one takes 8–10 minutes of live data entry. Each error creates rework that costs 3–5× the original call.

SKYLA handles every FNOL intake — compliantly, autonomously, in real time — from first word to filed claim and booked appointment.

EU AI Act Compliant GDPR Native DNB Ready Sub-200ms Multi-Tenant SaaS
The FNOL problem, quantified
— FNOL calls per agent, per day
8–10 min Average handling time per intake call
— Disconnected systems touched per call
3–5× Cost multiplier from a single data entry error
68% Customers say FNOL shapes their insurer satisfaction
The Problem with Existing Approaches

Every current solution fails FNOL
for the same structural reason.

❌ Fails Here
IVR & Traditional Voicebots
Rigid scripts break the moment a caller deviates. Cannot maintain context across a multi-step, emotionally charged conversation. Falls over on interruptions, corrections, and non-linear disclosures.
❌ Fails Here
Generic LLM Assistants
Probabilistic by nature — they guess. FNOL requires deterministic compliance: fraud checks must run, required fields must be validated, the order of steps is legally significant. LLMs cannot guarantee this without a constraint layer.
❌ Fails Here
CCaaS Platforms (e.g. Netcall)
Agent-assist tools that automate routing and summarisation. They still depend on a human agent to make judgement calls and execute across systems. The bottleneck is not routing — it's intake execution.
❌ Fails Here
RPA & Workflow Automation
Works for static, predictable flows. Breaks immediately when a caller interrupts, changes their account of events, or adds a third party mid-conversation. Live voice is fundamentally non-deterministic input.
SKYLA Agentic Voice AI

LLMs propose.
Deterministic logic decides.

SKYLA's Agentic Runtime Fabric (ARF) wraps AI inside policy-governed, symbolic rule enforcement. The LLM handles natural, empathetic conversation. The ARF ensures every step — fraud check, field validation, system write — executes in the correct order, with the correct data, under regulatory constraint.

⚡
Sub-200ms End-to-End Latency
Voice interface to confirmed system action under 200ms. No perceptible delay for the caller.
🛡
Six-Layer Hallucination Control
Pre-LLM PII filtering · Schema validation · OPA policy gates · Confidence gating · Execution-time validation · GER audit records. Nothing reaches a system write unless all six layers clear.
🏗
Multi-Tenant Architecture
Every insurer is a fully isolated tenant — credentials, policy rules, and data never cross boundaries. One deployment, N insurers.
🇪🇺
EU AI Act & GDPR by Architecture
Not a compliance checkbox — PII redaction runs before any LLM call. Every ARF invocation produces a Governed Execution Record ready for DNB audit.
Agentic Runtime Fabric — Layer Stack
Layer 1
Voice Interface
Telnyx WebRTC STT / TTS Interruption Handling
↓
Layer 2
Agentic Reasoning
LLM Routing Intent Engine RAG Context Memory
↓
Layer 3 — ARF Core
Policy & Constraint Enforcement
OPA Rules Schema Validation Confidence Gates Step Order
↓
Layer 4
Action & Integration
Claims System Fraud Check Scheduling CRM Write
↓
Layer 5
Observability & Compliance
GER Audit Log EU Data Residency Datadog
The ARF principle: LLMs are creative engines within deterministic guardrails — not decision-makers. Every system write requires policy clearance.
The FNOL Workflow

Five steps. Every call. Autonomously.

Every Dutch non-life FNOL intake follows the same five steps — today executed manually across four systems. SKYLA executes all five in a single, uninterrupted agent session.

Today — Manual
SKYLA — Autonomous
1
Today
Identity & Policy Verification
Agent searches policy admin system manually. Caller waits on hold during lookup. Errors in policy number entry cause re-verification loops.
SKYLA
Instant Identity Confirmation
SKYLA verifies identity via voice + policy lookup in real time. Caller continues speaking — no hold required. ARF gates step 2 until verification clears.
2
Today
Accident Detail Collection
Agent types structured fields during live call — date, location, third parties, circumstances. Interruptions cause data loss. Non-linear disclosures break data entry flow.
SKYLA
Structured Extraction, Live
SKYLA extracts all required fields from natural conversation, handles interruptions, corrections, and re-sequenced disclosures. Schema validation rejects invalid data before write.
3
Today
Fraud Flag Check
Agent must tab to fraud/risk system, manually enter policy number, interpret flag results mid-conversation. Step is sometimes skipped under call volume pressure.
SKYLA
Mandatory Automated Check
ARF mandates fraud check execution — it cannot be skipped. Results inform SKYLA's conversation routing. Flagged cases escalate to human agent with full context pre-loaded.
4
Today
Claim File Creation
Agent opens claims management system, manually populates all required fields. Data from steps 1–3 must be re-entered or copy-pasted. Entry errors create downstream rework.
SKYLA
Zero-Error Claim Write
SKYLA writes a complete, validated claim file atomically — all fields from steps 1–3 are carried forward. No re-entry. Schema-validated before write. Claim ID confirmed to caller live.
5
Today
Appointment Booking
Agent switches to scheduling system, checks assessor availability, proposes slots verbally, confirms with caller, then manually adds appointment. Step sometimes deferred to a callback.
SKYLA
Confirmed in the Same Call
SKYLA books assessor slot, confirms appointment to caller by voice and SMS — within the same call. No callback required. Intake is complete before the caller hangs up.
The Transformation

What changes when SKYLA runs your FNOL intake

Human agent navigates 4 systems manually during a live call, re-entering data at each step while keeping the caller engaged.

→

SKYLA executes all system actions autonomously, in compliant order, in real time — while maintaining natural, empathetic conversation with the caller.

Data entry errors during FNOL create rework downstream — incorrect claim files delay assessment, trigger compliance reviews, and require costly callbacks.

→

Schema validation before every write — no invalid claim files reach the queue. Error rate approaches zero by architecture, not by training.

Average handling time of 8–10 minutes. Agent burnout across 50–70 calls per day. Escalation capacity constrained by intake volume.

→

Intake handled end-to-end by SKYLA. Human agents handle edge cases and complex escalations only — the work that actually requires human judgement.

Integration Reality

Connects to your existing insurer IT landscape.
No custom builds. No rip-and-replace.

Standard insurer systems — pre-integrated
  • Guidewire ClaimCenter & PolicyCenter
  • Duck Creek Claims
  • Salesforce CRM
  • SAP Insurance
  • OpenText / Hyland OnBase
  • Telnyx & Twilio (voice carrier)
How connections work — AIS Registry
  • REST / OAuth2 via ARF generic executor
  • OpenAPI spec import → AIS auto-generated
  • JSONata field mapping — no code required
  • Per-tenant credential isolation
  • Sandbox regression tests before go-live
  • Configuration only — no development per insurer
2–4 wk
Onboarding timeline for standard Guidewire or Duck Creek environment
0 code
Changes required to add a new insurer — configuration only via AIS registry
100%
Tenant isolation — credentials, data, and policy rules fully separated per insurer
GDPR compliance is architectural: PII redaction runs before any LLM call. OPA-backed policy enforcement governs field-level data access. Every ARF invocation produces a Governed Execution Record (GER) — tenant ID, session, policy decision, outcome — ready for regulatory audit.
Why It Scales

Built for the Dutch market.
Repeatable across every insurer.

Every Dutch non-life insurer runs the same FNOL workflow, constrained by the same DNB regulations and GDPR requirements. What differs per insurer is their system landscape — not the workflow logic. SKYLA handles that variation through configuration, not custom development.

Pilot
Centraal Beheer
Full configuration · validation sandbox
→
Deployment
Insurer 2
New config layer only · 2–4 weeks
→
Scale
Insurer 3–N
Marginal cost decreases per tenant
→
Expansion
Adjacent Workflows
GGZ wachtlijst · Claims status · Renewals
Live Demo

Hear SKYLA handle
a real FNOL intake. Live.

No slides. No recordings. SKYLA will call you right now and walk you through a complete motor claims FNOL intake — identity verification, accident details, fraud check, claim creation, and appointment booking — as a real conversation.

1
Fill in your details — we capture this for follow-up and to personalise the demo call.
2
Click Start Demo Call — SKYLA connects via your browser (WebRTC, no app needed).
3
Speak naturally. Interrupt. Correct yourself. SKYLA adapts — and completes the full FNOL intake.

This demo runs on the same ARF infrastructure as production. Latency, compliance enforcement, and system integrations are live — not simulated. 🇳🇱 EU-hosted · GDPR compliant

Start FNOL Demo Call
Your browser will connect directly — no downloads required
Your details are used to personalise the demo and for follow-up only. We do not share your data with third parties. Privacy policy →
Connecting to SKYLA FNOL Demo Agent…
Call connected — SKYLA is ready. Speak naturally.
Connection failed. Please check your microphone permissions and try again.