Tokio Marine & Nichido Fire

Tokio Marine & Nichido Fire, one of Japan's largest P&C insurers, faced a compounding set of operational pressures in its claims division. Call center volumes were climbing as adjusters chased claimants for incomplete information submitted during first notice of loss (FNOL), a direct consequence of legacy data collection practices that captured insufficient detail at intake. Simultaneously, the insurer's fraud detection capabilities were limited to reactive, rule-based checks that missed sophisticated or slow-developing fraud patterns across a high volume of claims. The combined effect was rising loss adjustment expenses, mounting fraud-related losses, and deteriorating customer satisfaction — risks that threatened both profitability and competitive standing in a market where claims experience increasingly drives policyholder retention.

Tokio Marine & Nichido Fire deployed EIS ClaimSmart, a two-component platform purpose-built for P&C claims automation. The first component, ClaimPulse, modernized FNOL intake by delivering an omni-channel digital filing experience with 24/7 availability, structured data capture, and digital document uploads — replacing fragmented, manual collection workflows. The second component, ClaimGuard, applied a predictive machine learning risk-scoring model that continuously evaluates every claim in the portfolio, both newly filed and mature, rather than screening only at submission. The model's risk tolerances self-adjust as it ingests more data, improving precision over time. Integration with existing claims workflows allowed adjusters to act on scored outputs without overhauling downstream systems, enabling a practical deployment path for an enterprise-scale insurer.

The implementation produced measurable improvements across fraud containment, financials, and customer experience. The headline outcome was a 5x increase in fraud instances identified compared to pre-deployment methods — a result attributable to ClaimGuard's continuous monitoring of the full claims portfolio, not just newly submitted claims. The broader financial impact reached millions of dollars in annual savings, driven by reductions in both fraud-related losses and loss adjustment expenses stemming from more complete FNOL data. Customer-facing outcomes also improved materially:

• NPS spiked post-implementation, reflecting faster, less friction-heavy claims interactions
• Downstream rework declined as better intake data reduced the need for manual follow-ups
• Operational throughput improved as adjusters focused on flagged claims rather than reviewing all cases uniformly

• Continuous ML scoring across the entire claims portfolio — not just at FNOL — is what produced the 5x fraud detection lift; point-in-time checks at submission miss patterns that emerge as claims mature.
• Improving FNOL data quality has compounding returns: it reduces adjuster follow-up costs and simultaneously feeds better inputs to downstream ML models.
• Omni-channel digital intake is a prerequisite for fraud-scoring accuracy — garbage-in, garbage-out applies directly to risk models trained on claims data.
• Self-adjusting risk tolerances allowed the model to improve without manual recalibration, reducing the ongoing operational burden on the analytics team.
• NPS gains from claims digitization can offset the business case even before fraud savings are counted, making executive alignment easier for P&C carriers.