Solving the 'AI That Won't Answer Internal Matters' Problem – Launch of Enterprise RAG 'GMO AI RAG' [GMO Prime Strategy]

GMO Prime Strategy Co., Ltd. (President and CEO: Tadashi Yoshimasa, Head Office: Chiyoda-ku, Tokyo, hereinafter "GMO Prime Strategy"), a member of the GMO Internet Group, will begin offering the Enterprise edition of "GMO AI RAG," a RAG (Retrieval-Augmented Generation) software for corporations that instantly generates answers based on internal documents, starting Tuesday, June 30, 2026. An open-source version is also scheduled for release around Autumn 2026.

"GMO AI RAG" supports operation not only in cloud environments but also in on-premise/local environments, enabling the construction of a secure, custom AI knowledge base environment at low cost while minimizing information leakage risks. The Enterprise edition includes paid options and support services to enhance usability, maintainability, and performance. It strongly supports AI utilization for a wide range of corporations, from small and medium-sized enterprises to large corporations, from both cost and security perspectives.

[Background of Launch]

While the business use of generative AI is expanding, challenges such as "concerns about confidential information leakage," "AI chat services cannot answer questions about company-specific information," and "difficulty in grasping introduction and running costs" remain significant barriers in corporate operations. Especially for industries with strict security requirements and companies possessing vast amounts of documents and knowledge, building a system that allows for safe and efficient AI utilization of internal knowledge has become an urgent management issue.

Based on the technology cultivated through the development of the enterprise RAG platform "MAGATAMA Stack," GMO Prime Strategy, in combination with the sales network and brand power of the GMO Internet Group, will newly launch "GMO AI RAG" to deliver a custom AI utilization infrastructure to more companies.

[About "GMO AI RAG" (formerly MAGATAMA Stack)]

"GMO AI RAG" is software that ingests internal manuals, regulations, meeting minutes, proposal documents, etc., as knowledge and returns highly accurate answers based on internal information to inquiries from employees and AI agents. It also supports operation in a completely closed on-premise environment, allowing AI to be utilized for business while minimizing information leakage risks.

<Key Features>

● Transparency and Continuity through Open Source: Provided as OSS under the AGPL license (*1). It avoids vendor lock-in and enables highly transparent operation over the long term.

● Seamless Integration with Existing Storage: Supports integration with existing environments such as Microsoft SharePoint, Google Drive, Box, and Windows file servers. Knowledge can be utilized as-is without moving files. Please note that some integration functions include paid options or are under development.

● AI Optimized for Each Department: Equipped with an "Information Collection" function that segments information by department, such as sales, HR, and customer support. Access rights, LLMs used, and monthly token limits can be individually managed for each collection. Centralized control of access for the entire organization is achieved through integration with Microsoft Entra ID/Google Login.

● Ready for the AI Agent Era: Includes MCP (Model Context Protocol) server functionality as standard. It integrates with major AI chat services and AI agents such as Microsoft Copilot and Claude, enabling the utilization of internal knowledge from external tools.

● Flexible LLM (Large Language Model) and Infrastructure Configuration: The optimal LLM can be selected according to the use case, from utilizing major LLM services focused on performance (Anthropic, OpenAI, Google) to local LLM configurations that minimize information leakage risks. It supports a wide range of environments, from the latest cloud environments to completely closed on-premise environments.

(*1): AGPL (Affero General Public License) is a copyleft open-source license that may require source code disclosure even when providing services to third parties over a network.

[Paid Option for High-Accuracy Search "Tuned RAG"]

"GMO AI RAG" offers a paid option "Tuned RAG" to further enhance search accuracy.

Conventional RAG relies on simple vector search or keyword search, leading to issues such as missing proper nouns or paraphrased expressions, and failing to display highly relevant documents at the top.

This option pre-generates data compatible with multiple search methods during document registration and scrutinizes relevant documents in three stages: "candidate extraction," "re-evaluation," and "final ranking" during questioning. This makes it easier to find documents with different expressions or contextually related information, allowing for more accurate retrieval of information necessary for answer generation. Furthermore, when fragmenting documents for search, it ensures context is maintained by adjusting fragments using neural algorithms to prevent unnatural breaks in meaning.

In terms of F1 score (*2), a measure of search accuracy, this option achieved approximately 0.591, compared to approximately 0.332 to 0.417 for existing search algorithms. This is approximately 1.4 to 1.8 times that of existing search algorithms, confirming that necessary information can be searched with higher accuracy.

(*2): F1 Score: A metric for measuring search accuracy. It evaluates how well the system balances "the degree to which necessary documents are not missed (recall)" and "the degree to which irrelevant documents are not retrieved (precision)" on a scale of 0 to 1. A higher score closer to 1 indicates higher accuracy.

(*3): JQaRA (Japanese Question Answering with Retrieval Augmentation): A dataset constructed to evaluate "Japanese search accuracy" in AI retrieval-augmented generation (RAG) systems.

Figure: Comparison of F1@10 values for JQaRA (*3) across all 1,668 questions.

[Main Pricing Plans]

* OSS version scheduled for release around Autumn 2026.

* The above are reference prices (excluding tax). Costs will vary depending on the scope of implementation support, number of users, support plan, etc. Infrastructure costs such as AWS or cloud LLM usage fees and hardware expenses are separate.

[Intended Use Cases]

● Streamlining Internal Help Desks and Back Offices: Knowledge base created from employment regulations, internal rules, and various procedure manuals. AI automatically handles routine inquiries to IT departments and general affairs/HR, significantly reducing the burden on front-line operations.

● Accelerating Sales Activities and Field Support: AI instantly searches and summarizes vast product specifications, past proposal documents, and success stories. Improves response speed to customers and supports sales representatives in standardizing skills and increasing closing rates.

● Technical Succession and Asset Creation of Tacit Knowledge: Consolidates work procedure manuals and past troubleshooting histories of veteran engineers into AI. Creates an environment where junior engineers can access knowledge in real-time on-site, inheriting valuable know-how as organizational assets.

● Speeding Up Legal and Compliance Responses: Knowledge base created from contract templates, legal interpretation documents, and internal compliance guidelines. Provides an environment where personnel can perform initial checks without consulting the legal department, simultaneously enhancing decision-making speed and risk management levels.

● Improving Customer Support Quality and Standardizing Responses: AI performs cross-searches of product FAQs, past inquiry response histories, and complaint cases. Operators can instantly refer to the basis for their answers, reducing variations in response quality, improving customer satisfaction, and lowering training costs.

Product Introduction Site: https://ai-prime-strategy.gmo/lp/gmo-ai-rag/

[Future Developments]

Going forward, we will accelerate functional enhancements such as API integration with major SaaS, "Remote MCP" that allows referencing data within Gmail and Slack as knowledge, and GPU server support to strengthen local LLM environments. We will continue to adapt to the latest LLM and AI utilization trends, and under the slogan "Secure AI for All Companies," we will provide a platform that maximizes competitiveness and productivity while protecting companies' critical information, thereby strongly promoting the social implementation of generative AI.

[About GMO Prime Strategy]

GMO Prime Strategy, with its corporate philosophy of "Everything for the Development of the Enterprise OSS Ecosystem," aims to solve social issues through technology-centric mechanisms and create social value.

It develops the web acceleration solution "KUSANAGI Stack," which is offered through major cloud providers and other platforms, and provides full-managed support services "KUSANAGI Managed Service" for integrated maintenance and management of CMS including WordPress, and servers, as well as cloud integration services and licensing services.

End

[Inquiries Regarding This Solution]

● GMO Prime Strategy Co., Ltd.

AI Business Division

https://www.prime-strategy.co.jp/contact-us/

[GMO Prime Strategy Co., Ltd.] (URL: https://www.prime-strategy.co.jp/)

Company Name: GMO Prime Strategy Co., Ltd. (TSE Standard, Securities Code: 5250)

Location: 1st Floor, Sumitomo Fudosan Ichibancho Building, 1-8 Ichibancho, Chiyoda-ku, Tokyo

Representative: Tadashi Yoshimasa, President and CEO

Business Activities: KUSANAGI Stack Business

- Managed Services

- Cloud Integration Services

- Licenses

Development and Provision of KUSANAGI Stack

- Ultra-high-speed CMS execution environment "KUSANAGI"

- High-speed engine "WEXAL® Page Speed Technology®"

- Strategic AI "ONIMARU® David"

Capital: 453,798 thousand yen (as of the end of November 2025)

[GMO Internet Group, Inc.] (URL: https://group.gmo/)

Company Name: GMO Internet Group, Inc. (TSE Prime, Securities Code: 9449)

Location: Cerulean Tower, 26-1 Sakuragaoka-cho, Shibuya-ku, Tokyo

Representative: Masatoshi Kumagai, Representative Director, Group Representative

Business Activities: Holding Company (Group Management Function)

Internet Infrastructure Business

Internet Security Business

Internet Advertising and Media Business

Internet Financial Business

Cryptoasset Business

Capital: 5 billion yen

* Company names, product names, etc., mentioned herein are trademarks or registered trademarks of their respective companies, registered in the U.S. and other countries.

Copyright (C) 2026 GMO Prime Strategy Co., Ltd. All Rights Reserved.