Reinsurer Underwriting Assist
A reinsurer deployed an underwriting assist agent grounded in a policy knowledge graph — passed model risk review with explainable recommendations and reduced analyst research time by 50%.
A reinsurer deployed an underwriting assist agent grounded in a policy knowledge graph — passed model risk review with explainable recommendations and reduced analyst research time by 50%.
Last updated:
Challenge
Underwriters spent hours synthesising treaty terms, exposure data, and historical loss patterns — while an internal LLM pilot produced confident but ungrounded summaries unsuitable for model risk submission.
Production deployment — not proof-of-concept
Governed agents with full audit trails
Embedded FDE delivery model
Operational accountability through go-live
Approach
Knowledge graph structured treaty entities, exclusions, and exposure linkages. Agent retrieved governed context and produced recommendations with mandatory citations.
Model risk documentation included eval results, limitations, and human-in-the-loop approval for bind/no-bind recommendations.
Results
50% research time reduction. First-pass model risk approval. Underwriters retained final authority with agent-prepared briefing packs.
Production outcomes, not pilot metrics.
Underwriter research time on in-scope reviews.
Recommendations with policy citations.
Model risk committee approval.
Related resources
Ready for an AI implementation partner?
Book a discovery call and we'll map your highest-value use case — and exactly how we get it into production.
Frequently asked questions
- Can Derisk360 provide references?
- Selected references are available on request during discovery calls.
- What industries are represented?
- Banking, insurance, financial services, and regulated enterprises globally.
- How do I start a similar engagement?
- Book at derisk360.com/book.