Annual Meetings-3A. How AI is Evolving Inland Marine Risk

3A. How AI is Evolving Inland Marine Risk - Jay Bell

Pdf Summary

This presentation argues that AI literacy for underwriters is less about technical knowledge and more about using AI to improve judgment, efficiency, and decision quality. It opens by describing a familiar underwriting reality: submissions arrive with missing or unclear information, brokers expect fast responses, and poor decisions can lead to bad pricing or bad risk selection.<br /><br />The core message is that AI’s main value is not just speed, though it can quickly summarize documents, draft communications, and do first-pass reviews. The bigger opportunity is better decision-making: asking better questions earlier, organizing complex risk information, improving file quality, and increasing confidence in underwriting decisions.<br /><br />The presentation explains generative AI as a prediction system that relies heavily on context. Like a guessing game with missing letters, vague prompts produce vague outputs, while detailed, structured context leads to better results. The central insight is: better context leads to better decisions, both in AI and underwriting.<br /><br />AI is positioned as a tool for professionals, not a replacement for them. It is useful for synthesis, drafting, pattern recognition, and surfacing gaps, but it is weak at nuanced judgment, risk appetite decisions, and detecting confident-sounding errors. Underwriters should direct AI with clear intention: define the decision, constraints, evidence, output format, and verification steps.<br /><br />Practical prompt patterns include triaging submissions, comparing documents, and red-teaming assumptions. The recommended habit loop is: draft, critique, verify, then finalize. The conclusion is that AI should extend underwriting expertise, not replace it, and responsible adoption depends on clarity, discipline, and human judgment.

Keywords

AI literacy

underwriting

decision quality

generative AI

prompt engineering

risk assessment

context

judgment

efficiency

human oversight