This work formalizes a vertically integrated AI paradigm for motor insurance, combining structured visual perception, contextual reasoning about vehicle components, and multimodal document intelligence into a unified framework. The handbook traces the evolution of MARSAIL from experimental prototypes to a fully operational ecosystem, detailing sequential innovations including MARS for hierarchical spatial refinement, ALBERT for global contextual reasoning, SLICK for efficient knowledge-distilled deployment, and DOTA for robust document understanding. Each contribution addresses a critical limitation in existing AI systems and collectively establishes a production-ready pipeline for end-to-end automation of vehicle damage analysis, claims evaluation, and underwriting workflows. Designed as both a technical reference and a research blueprint, the handbook demonstrates how domain-specific AI architectures, aligned with MLOps and operational constraints, can achieve interpretability, reliability, and scalability, offering a definitive guide for researchers and practitioners seeking to build AI-driven solutions in motor insurance.
