Lomin, a Document AI company led by CEO Kang Ji-hong, announced on September 9 that it has supplied its Document AI Agent, Lomin, to Korean Reinsurance Company's project to build an AI assistant for specialty insurance rating.

The project was launched to apply AI to rating work in the specialty insurance underwriting process. Underwriters previously reviewed a range of documents, such as rating requests (RQs) and questionnaires, looked up historical data on their own to set rates, and entered the related information by hand. With the AI assistant in place, the entire process, from document intake and information extraction to searching and referencing historical data and setting rates, is now automated.

In this project, Lomin handled RAG (retrieval-augmented generation) preprocessing for the rating AI assistant. It classified and parsed incoming documents such as rating requests and questionnaires, extracted the required information, and converted it into structured data.

For rating requests that span multiple pages, Lomin developed and applied a multi-page document extraction AI model. The model automatically extracts and consolidates information scattered across pages, so underwriters no longer have to check and compile it item by item. Lomin also converted past submissions into data so the AI assistant can search and reference them alongside new submissions.

Lomin is a Document AI Agent that processes a wide range of unstructured documents with a VLM (vision language model) and automates document work by connecting to enterprise business systems. Its grounding feature shows where each result comes from in the source document, and field-level confidence scores let reviewers check only the items that need attention.

"To apply generative AI and AI agents to real work in insurance and other financial sectors, the key is turning complex and diverse documents into accurate data," said Kang Ji-hong, CEO of Lomin. "Building on our experience converting documents into data for large-scale AI projects at major insurers, banks, and securities firms, we will lead the Document AI field for generative AI and AI agents in finance."

Lomin has carried out a series of projects in the financial sector to build the document data that generative AI relies on. It took part in generative AI projects at KB Financial Group and Samsung Securities, and it is currently responsible for recognizing and structuring 150 types of business documents in Woori Bank's AI agent banking project. It has also supplied document processing systems to insurers including Kyobo Life, Lina Life, and Heungkuk Life.