Industry First: Generative AI Used for Medical Judgment in Benefit Payment Assessment; Cinnamon AI to Fully Deploy "Benefit Payment Assessment AI" for Daido Life Insurance

Cinnamon Inc. (Headquarters: Chiyoda-ku, Tokyo; Representative Director and CEO: Miku Hirano; hereinafter "Cinnamon AI"), a provider of AI solutions to major domestic companies, will fully deploy its "Benefit Payment Assessment AI" starting April 1, 2026, for Daido Life Insurance Company (President: Mutsuro Kitahara; hereinafter "Daido Life"), a member of the T&D Insurance Group. This marks the first time in the life insurance industry (*1) that generative AI and RAG (*2) have been utilized for medical judgment in benefit payment assessment operations.

In the process of claiming hospitalization benefits for life insurance, there is a step where the necessity of aggregating hospitalization days must be determined based on medical judgment regarding the diseases involved in past and new claims. In cases where this judgment is difficult, assessment staff have traditionally had to consult with in-house doctors, which has been a challenge due to the significant costs associated with preparation and waiting for responses.

Cinnamon AI has now co-developed (*3) an AI agent that utilizes generative AI with RAG to highly effectively leverage the business manuals and medical information held by Daido Life. With the introduction of this solution, the AI can now generate medically grounded judgments in place of doctors for approximately 70% of cases that previously required internal medical review, doing so accurately and in a short amount of time. This improves the processing speed per assessment, shortens the number of days required for payments to customers, and is expected to reduce the workload of assessment staff and in-house doctors by approximately 1,800 hours annually. By reallocating this saved time to the study and implementation of administrative processes that create optimal customer experiences, personnel can be shifted to higher value-added tasks, further promoting DX across the entire organization. Additionally, by having less experienced assessment staff reference and learn from the medical judgment grounds output by the generative AI, they can efficiently acquire medical knowledge, which is expected to further improve assessment quality.

Cinnamon AI intends to continue deepening its partnership with Daido Life, expanding the application areas of the AI agent, and applying the verification and analysis results of AI-driven judgments to review conditions for automated processing, aiming for further efficiency and speed in payment assessment operations.

*1 According to research by Daido Life Insurance Company. *2 Retrieval Augmented Generation: A technology that adds proprietary information, such as internal documents, to large language models. *3 The AI solution developed this time was created by Daido Life Insurance Company, Cinnamon Inc., and DiCE Inc.

Cinnamon AI is currently providing two AI products that handle the core of business operations

◇ Super RAG™ "Super RAG" is a proprietary AI product that can build an LLM environment capable of analyzing complex documents, such as tables and charts, and generating highly accurate answers without the need for tuning. Its greatest feature is the "overall optimization" of the entire process—preprocessing, retrieval, and generation—in a versatile and high-precision manner. Unlike conventional RAG, which mechanically segments text by character count, the preprocessing stage uses proprietary layout analysis technology to automatically identify titles, charts, and tables, optimally chunking the document while preserving its meaning and structure. This highly structures unstructured data that is otherwise unsuitable for data utilization, enabling business automation via RAG systems with extremely high precision. https://cinnamon.ai/super-rag/

◇ Flax Scanner HUB "Flax Scanner HUB" is a highly versatile, original, high-precision AI-OCR platform that allows AI to understand meaning and read data from documents of various formats used across diverse industries and sectors, just as a human would. Conventional AI-OCR has primarily relied on "coordinate definition" types that require pre-definition for formatted documents, making it difficult to handle documents with countless formats. In contrast, "Flax Scanner HUB" can extract data with high precision even from unstructured forms that are difficult to read, without complex prior settings. It is equipped with three different AI engines—coordinate definition, feature learning, and generative AI extraction—and reads data using the optimal method for the type of form. https://cinnamon.ai/flax-scanner-hub/

* "Super RAG™" and "Flax Scanner™" mentioned in this press release are trademarks of Cinnamon Inc.

Cinnamon AI Overview Company Name: Cinnamon Inc. URL: http://www.cinnamon.ai Location: 6F Kaneda Building, 3-16 Kioicho, Chiyoda-ku, Tokyo Established: October 2016 Representative: Miku Hirano, Representative Director and CEO

[Business Description] AI product business targeting unstructured data, AI consulting business

With the purpose of "creating a world full of creativity where everyone can envision a new future, together with AI," Cinnamon AI provides AI products and AI consulting to realize advanced business AI solutions. By combining overwhelming high-precision document analysis technology with large language models (LLMs), we promote the "automation of automation." Centered on the innovative AI-OCR "Flax Scanner," which can integrate generative AI, and "Super RAG," which significantly promotes LLM utilization, we act as a leader in transformation, helping customers create new business structures in the AI era from their existing knowledge and achieve exponential growth. We have a proven track record of providing services to numerous major domestic companies. We have established an artificial intelligence research laboratory in Vietnam (Hanoi and Ho Chi Minh City) to build an ecosystem for acquiring and nurturing high-level AI talent, accumulating technical assets that utilize unstructured data, such as AI-OCR, natural language processing, large language models, and speech recognition. Together with our unique AI strategy, the "Harvest Loop," we use these technologies to support the promotion of DX and AI utilization.