Fraud Detection · Risk Management
Zurich needed to review very large claims volumes more efficiently, reduce investigator drudgery and false positives, and detect fraudulent claims and emerging fraud patterns quickly enough to protect honest customers and contain unnecessary claims costs.
Reported impact
Zurich says the system reduces false positives, increases the number of suspicious and confirmed fraud cases found, helps the company spot fraud trends earlier, and speeds differentiation between honest and fraudulent claims so legitimate claims can be processed faster. In Brazil, Zurich reports that the fraud detection rate in the affinities business trebled after adopting ZCAM’s system.
Zurich Insurance Group runs an in-house AI-assisted fraud detection workflow for claims investigations. In Switzerland, incoming non-life claims from the Allegro core claims system are funneled into Zurich’s proprietary fraud scoring tool, which assigns each claim a 0 to 100 score and flags higher-risk cases for the Special Investigation Unit. The scoring process combines rules with a machine learning model that Zurich describes as natural-language-based generative AI trained to detect words commonly present in fraud cases, initially piloted in German for property and commercial lines. Investigators also use Zurich’s internal AI chatbot in daily work. The solution was built by Zurich Customer Active Management (ZCAM), whose specialists tune the models and workflows with claims teams. The article says claims data is stored using cloud computing, but does not name the storage platform. Outputs are risk scores, anomaly signals, fraud trend insights, and faster routing of suspicious versus honest claims.
Claims entered in Allegro are funneled into Zurich Switzerland’s proprietary fraud scoring tool, where rules and a machine learning NLP model analyze claim details and assign a 0 to 100 score. Higher-scoring cases are escalated to the Special Investigation Unit for human investigation, while investigators also use Zurich’s internal AI chatbot in daily work and insights from the system are used to refine future fraud indicators and trend detection.
"Along with a rule-based system, to gain more accuracy and efficiency, Zurich’s Swiss Special Investigation Unit is now testing a machine learning model, a form of generative AI based on natural language processing"
"Besides developing IT solutions in-house, Zurich’s Swiss Special Investigation Unit is also exploring ways to use new technology, including incorporating Zurich’s internal AI chatbot in daily work to support investigations."
"It’s about reduced false positives," says Andrew Fairclough
"Most of the more than 600,000 claims it gets annually, with the exception of life insurance claims, are funneled through the system, and given a score from 0 to 100."
"The fraud detection rate in the affinities business, for example, has trebled since Zurich Brazil adopted ZCAM’s system"
"With our systems, paired with humans, we can better, and more quickly differentiate between honest and fraudulent claims. That should allow honest customers to get their claims processed quickly"
"Zurich’s AI is being trained to spot certain words often present in fraud cases."
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
In-house Sample Sep 24, 2026 Mastercard Fraudsters are using AI to find vulnerabilities, adapt tactics, and scale attacks across more payment channels, while traditional transaction-by-transaction controls struggle to catch coordinated fraud and authorized… Last year, scam losses globally were estimated at $442 billion In-house Sample Feb 1, 2024 Mastercard Mastercard needed to improve fraud detection for banks and cardholders as fraudsters increasingly exploit technology, while also reducing false positives that incorrectly flag legitimate transactions. 143 billion transactions a year Decision Intelligence · Decision Intelligence Pro In-house Sample Jun 25, 2026 Telefónica Telefónica's compliance function must manage an expanding set of digital risks and increasingly complex international regulations without becoming inefficient, while also improving fraud detection and preserving rights… 