Outcreating Underwriting Data Lag in P&C Insurance with Agentic AI on Amazon Bedrock

Our Client

The client is a leading US insurer with more than 160 years of experience and one of the largest writers of commercial and personal property and casualty insurance in the United States. Working through independent agents, it provides coverage for a wide range of needs, including auto, home, renters, and business insurance.

Its Bond and Specialty Insurance line of business relies on high-quality external data to price risk accurately and remain competitive.

Industry Trends

P&C insurers are using generative AI in insurance underwriting to make decisions faster, more data-driven, and more resilient. Leading carriers are increasingly:

• Using generative AI to ingest and interpret external and third-party data
• Deploying agentic AI in insurance to answer underwriter questions in natural language
• Standardizing data from diverse public sources to feed risk and pricing models
• Shortening quote-to-bind cycles to win more business
• Building on secure, enterprise-grade cloud AI platforms such as Amazon Bedrock

Challenges

The insurer's underwriters faced several challenges in working with external data.

• Dependence on external data: Bond and Specialty underwriters rely on external data sources, such as data published by agencies including the labor department, for underwriting and risk management.
• Constantly changing data: External data changed over time due to modifications in how it was captured, changes in collection methodologies, and variations in granularity as data was rolled up or broken down.
• High stakes for accuracy: This data feeds multiple risk and underwriting models, so incorrect translation or preparation could lead to inaccurate underwriting, flawed risk ratings, and reduced market competitiveness.
• Slow adaptation: Historically, adapting to these data changes took several months, slowing responsiveness and decision-making.

These challenges delayed underwriting decisions and limited the insurer's ability to respond quickly to market opportunities, underscoring the need for generative AI in insurance underwriting.

LTM's Solution

LTM developed the GenAI Event Analytics (EA) Bot, an application of agentic AI in insurance powered by Anthropic's Claude 3.5 Sonnet on Amazon Bedrock.

• Prompt generalization and attribute standardization: LTM created algorithms that generalize prompts and standardize attributes, enabling consistent, reliable responses from Claude 3.5 Sonnet on Amazon Bedrock even as source data changes.
• LLM-powered event data parsing: The solution captures medium to highly complex event data and uses Amazon Bedrock LLMs to parse it into usable information.
• Agentic AI query application: An agentic AI application lets users query data related to employment, construction, and health and safety incidents in natural language.
• Easy-to-consume outputs: The bot generates clear, tabular responses, removing the pain of analyzing raw external data and speeding up the assimilation of information needed to quote policies.

Business Benefits

By applying generative AI in insurance underwriting, the GenAI EA Bot delivered measurable improvements in speed and quality.

• Increased underwriting efficiency with near real-time data available to actuaries, underwriters, and field agents, compared with one to two months without generative AI.
• Reduced analysis time for business analysts by 25%, enabling faster time to market for bond and policy issuance.
• Improved risk selection through more accurate, consistently prepared external data.
• Enhanced overall customer experience and Net Promoter Score (NPS) through faster, more responsive underwriting.

Conclusion

With agentic AI in insurance underwriting, built on Amazon Bedrock, transforming how external data reaches underwriters, LTM helped the insurer replace months of manual data adaptation with near real-time, decision-ready insight. The GenAI EA Bot enables faster quoting and issuance, better risk selection, and a more responsive experience for agents and customers, opening new roads to growth in the Bond and Specialty business.

Ready to Outcreate underwriting data lag and AI-powered underwriting insights?

Contact us at info@ltm.com