Visada.it offers DAFNE, an AI-powered fraud detection platform for insurance claims that analyzes images, metadata, and documents to uncover anomalies and repeated damage patterns. The system anonymizes sensitive data, cross-references claims against company archives, and produces validated evidence reports for antifraud teams. It also detects AI-generated or edited images and flags inconsistencies in claim documentation.
Funding
Funding not disclosed
Founders
Product
Problem
Insurance companies face challenges in detecting fraudulent claims, as fraudsters reuse images, manipulate metadata, and submit forged documents that are difficult to identify through manual review. Traditional verification methods struggle to uncover patterns across claims that may be separated by time or geography, allowing sophisticated fraud schemes to go undetected.
Solution
DAFNE is an AI-powered platform that automatically analyzes insurance claims by examining images, metadata, and attached documents to identify signs of fraud. The system anonymizes sensitive elements like license plates and faces, then isolates and analyzes damage to build a technical representation that can be compared over time. It cross-references claims against the company's archive to reveal recurrences and reuse patterns, while also detecting discrepancies in metadata, editing traces, and AI-generated content. High-confidence matches are verified by specialized analysts to eliminate false positives, and confirmed cases produce structured evidence reports with direct comparisons and traceable references. The platform integrates four analysis levels that work together, with results only transmitted to the company after expert validation.
Target Audience
Primary customers are insurance companies' antifraud teams and claims departments that need to verify incoming claims and identify potentially fraudulent cases before settlement.
Features
- Automated image analysis that isolates damage, assesses shape, extent, and type, and creates comparable technical representations
- Metadata forensics detecting discrepancies between photo capture dates and claim dates, editing traces, and AI-generated or retouched images
- Document anomaly detection identifying modified headers, fictitious names and contacts, recurring phone numbers across unrelated claims, and reused documents with minor variations
- Cross-claim archive comparison that surfaces damage recurrences and reuse patterns across claims distant in time
- Human-in-the-loop validation process where specialized analysts review high-confidence matches to eliminate false positives
- Partnership with LambdAI Space extending analysis to properties, roofs, and external areas for property insurance claims