01
Niche insurer · Basque Country · 2,400 claims/month
Multi-channel claim processing
Intake from 3 channels, extraction with human validation, escalation when confidence <90%. NPS +11 pts in segment.
55 min → 27 min per claim
From 72 h to 6 h. First response to renewing policyholders.
We automate claim intake, quotes, lead qualification and processing with AI. Multi-channel ingestion, data extraction with human validation, and response-time KPIs from day one.
In insurance every extra hour of response impacts annual renewal. Manual operations is killing retention.
Email, portal, phone, WhatsApp. Each channel with different structure. Technicians lose 55 min per claim just classifying and transcribing.
Brokerages with high per-lead conversion costs. Response speed defines capture ratio.
DORA, IDD, Solvency II. Manual compliance scales poorly with volume.
Not a list of technologies. Concrete processes where AI delivers measurable euros or hours from the first month.
Extraction of key fields from email, portal, workshop PDFs, attached photos and transcribed calls. Routing to correct handler with measured confidence.
Chatbot or voicebot that qualifies inbound leads, gets minimum data and books with human intermediary in under 3 minutes.
Quote generation from client and risk data. The intermediary validates and signs; AI does the data entry and condition calculation.
Analysis of suspicious patterns: repeated claims, atypical amounts, discrepancies between claim and attachments. Alert to fraud team.
Generation of DORA, Solvency II and internal KPI reports with data consolidated from multiple management systems.
Voicebot for frequent queries (claim status, next premium, data change). Escalation to human when out of script.
Multi-channel data extraction, classification and routing with human validation.
Info → 02Automated first line with smart escalation.
Info → 03Business-line dashboards, loss ratios, actuarial forecasts.
Info → 04Fraud detection, policyholder data protection, DORA compliance.
Info →Invented cases with metrics consistent with our real ranges, until a client authorises publishing their own.
Intake from 3 channels, extraction with human validation, escalation when confidence <90%. NPS +11 pts in segment.
Voicebot 24/7 qualifies and books with correct agent. Lead-to-quote conversion +18%.
Models cross-reference history, patterns and attachments. Prioritised alert to fraud team for suspicious cases.
Our pilots incorporate from design the regulatory requirements that apply to the insurance sector in Spain:
We work from design with DORA (digital operational resilience) and Solvency II (governance and risk management) requirements. Every AI agent we deploy has dependency documentation, fallback plan if the model fails, and audit log satisfying CNMV/DGSFP in an inspection.
For DORA operational risks we classify each pipeline component (model, orchestrator, knowledge base) by criticality. Critical components deploy in your infrastructure, not external SaaS.
The pilot doesn't slow: we build regulatory evidence in parallel to development, not as a later phase.
Each agent has configurable confidence threshold. By default, any decision with confidence <90% escalates to human automatically. In production, escalation rate typically sits at 15-25% in the first 3 months and drops to 8-15% once the model learns.
Automated decisions are audited by random sample by your quality team. If we detect model drift (bias, pattern change), we retrain it.
On irreversible processes (payments, policyholder notifications, coverage denial) there's always human in the loop. AI proposes, human validates and signs.
Yes. Our stack connects via API to the main systems in the Spanish sector: Segurpro, GIS, Sifra, Zurich Advance, large insurers' proprietary platforms and internal ERPs. For legacy systems without API we use RPA integration + human fallback.
Integration is designed in week 2 of the pilot (technical design phase). Before signing, we verify which endpoints are available and what requires workaround to avoid implementation surprises.
Yes, that's how we recommend starting. An acotated pilot for a single line (motor, home, health) or a single claim type (property damage, liability, theft) validates technology, process and ROI in 6-8 weeks.
Then it scales to other lines with your own validated data. This incremental approach reduces the risk vs the big-bang deployment that usually fails in insurers due to organisational complexity.
With one line properly automated, you already recover the pilot cost in 3-5 months.
Free 30-min diagnosis. We come out with 3-5 use cases prioritised by ROI for your specific insurer.