This Hong Kong insurance company, one of the top 15 general insurers by gross written premium in the SAR, processed approximately 145,000 claims annually across motor, property, travel, and health insurance lines. The claims department employed 92 assessors who manually reviewed every claim — examining policy terms, verifying documentation, assessing damage reports, and making settlement decisions. The average time from claim submission to settlement was 23 business days, well above the company's target of 10 days.
Customer complaints about slow claims processing had increased by 40% year-over-year, and the company's Net Promoter Score had dropped from 34 to 18. The Insurance Authority had flagged the company's claims handling times in their supervisory review. Meanwhile, insurtech competitors were advertising 48-hour claims settlements, putting extreme competitive pressure on the company's retention rates.
Fraud was another significant challenge. The company estimated that 8-12% of claims contained some element of fraud, from exaggerated amounts to entirely fabricated incidents, but the manual review process only identified approximately 3% as fraudulent. The company's Special Investigations Unit of four analysts could only deep-dive into a fraction of suspicious claims, meaning potentially HKD 85 million in fraudulent claims were being paid annually.
Pertama Partners' AI Readiness Audit mapped the entire claims lifecycle from first notification of loss through to settlement and identified that 62% of claims were straightforward, low-complexity cases that followed predictable patterns — ideal for AI-assisted straight-through processing. We analyzed three years of claims data, including 435,000 historical claims with their outcomes, processing notes, and fraud investigation results.
Our AI Pilot Program built a three-tier claims processing system. The first tier used AI to automatically validate claim documentation, extract key information from photos and documents using computer vision and OCR, verify policy coverage, and calculate settlement amounts for straightforward claims — enabling same-day processing with human spot-check oversight. The second tier handled moderate-complexity claims by pre-populating assessor screens with AI-extracted information and settlement recommendations, reducing manual work by 60%. The third tier flagged high-complexity and potentially fraudulent claims for senior assessor review, with AI-generated risk scores and anomaly explanations.
The fraud detection component used network analysis to identify claim rings, anomaly detection to spot unusual patterns, and natural language processing to flag inconsistencies in claim narratives. AI Governance Retainer engagement established model monitoring, bias testing, and regulatory reporting aligned with IA guidelines. Team Training transformed claims assessors from data entry clerks into AI-augmented decision-makers focused on the cases that genuinely required human judgment.
"Pertama Partners helped us see that fast claims processing and rigorous fraud detection are not competing priorities — with AI, they reinforce each other. Our honest customers get paid faster, and fraudsters get caught more often."— Angela Leung, Chief Claims Officer
Every transformation starts with a conversation. Let's discuss your challenges and opportunities.
Discuss Your Challenge