Structure Unstructured Claims Data for Modeling Inputs

Actuarial models (pricing, reserving) typically require structured data inputs. However, valuable predictive information often resides in unstructured claims descriptions or adjuster notes. LLMs can process this free-text data, classify claims based on narrative details (e.g., injury type, cause of loss), or extract specific features (e.g., presence of litigation indicators) that can then be used as structured inputs to enhance traditional actuarial models.