Macnica signs Japan's first sales agency agreement with Unstructured, a company that automates the preparation of "unstructured data," a major obstacle to the practical application of RAG (Radio Aggregation). This will eliminate reliance on individual expertise, reduce operational burden, and support companies in utilizing AI in their own unique ways.

Macnica (Headquarters: Yokohama City, Kanagawa Prefecture, Representative Director and President: Kazumasa Hara, hereinafter referred to as Macnica) announced today that it has concluded a sales agency agreement with Unstructured Technologies, Inc. (Headquarters: California, CEO: Brian Raymond, hereinafter referred to as Unstructured) to become the first distributor in Japan, and will begin offering "Unstructured," a platform that automatically prepares unstructured data into a format that is easy for LLM (Large-Scale Language Model) to handle.

■Challenges in RAG adoption and unstructured data development
In recent years, with the spread of LLMs, expectations for knowledge retrieval and business efficiency improvements using generative AI have increased, and more companies are working on building RAGs (Search Enhancement Generators) using internal documents.
On the other hand, unstructured data such as sales materials, contracts, manuals, and technical documents within companies often have diverse document formats, making it difficult for AI to accurately grasp structures such as paragraphs, headings, tables, and images. As a result, RAG (Data Aggregation) experiences search omissions and misinterpretations of context, and issues such as unstable accuracy in production environments, even if it works fine in proof-of-concept (PoC), have become apparent.
Therefore, in building a RAG (Regional Aggregation System), it is essential to prepare this unstructured data into a format suitable for AI utilization. However, in typical data preparation processes, designers need to design and adjust chunking and information extraction rules for each document, which requires a high level of expertise and is prone to becoming dependent on specific individuals. Furthermore, readjustments are required every time the volume of documents increases or the content is updated, so the workload in the operational phase continuously increases. As a result, the burden of data preparation and operation, rather than the performance of the AI model itself, becomes the bottleneck, and in many cases, the full-scale deployment of generative AI does not occur, or the project stagnates.

■An unstructured approach to solving problems
Unstructured is a platform that automatically prepares unstructured data into a format that is easy for LLM to handle. By converting paragraphs, headings, tables, images, and other structural elements within a single document to JSON*, it contributes to improving the accuracy and stability of RAG.
This significantly reduces the time and effort previously spent by design engineers on detailed tuning for each document and on updating documents, thereby eliminating reliance on individual expertise and easing the ongoing operational burden.

[Point 1] Equipped with connectors for integration with various cloud services as standard.
It integrates with various cloud services using standard connectors, enabling continuous processing, including updates, of unstructured data without moving or replicating it to other storage.

[Point 2] Preprocessing design with the goal of "utilizing RAG and generation AI"
By using partitioning that considers document structure, chunking that preserves meaning, and metadata addition that anticipates subsequent AI processing, we transform unstructured data into a format that can be directly used in business operations.

[Point 3] Rapid PoC and implementation that can be started without coding.
It features a no-code GUI, allowing even users without specialized knowledge to reliably perform complex unstructured data processing.

[Point 4] Compliance with compliance and security standards
We comply with data protection and information security-related laws, regulations, and industry standards, including HIPAA, SOC 2 Type 2, GDPR, and ISO 27001.

<Achieving end-to-end data preparation with Unstructured>

Going forward, Macnica will accelerate the practical application of generative AI in Japanese companies by providing comprehensive support, from designing integrations with existing data and content management infrastructures to implementation.

In making this announcement, Brian Raymond, CEO of Unstructured Technologies, Inc., stated the following:
"Approximately 80% of corporate data worldwide is unstructured, buried within PDFs, emails, presentations, and various documents, and not being fully utilized by AI systems. This is the biggest bottleneck hindering the adoption of generative AI in enterprises, and a common challenge faced by many organizations across industries. Macnica, with its deep understanding of the Japanese enterprise market and strong commitment to cutting-edge AI solutions, is the ideal partner for deploying Unstructured in Japan. Through our collaboration, we will help Japanese companies unlock the true value of their data and move from proof-of-concept (PoC) of generative AI to large-scale deployment in production environments."

* JSON (short for JavaScript Object Notation) is a widely used data description format that represents data using a combination of "item names" and "values" across systems.

[Click here for product details]
URL: https://www.macnica.co.jp/business/dx/manufacturers/unstructured/

[For product inquiries, please contact us here]
Macnica Unstructured Products Team
E-mail: unstructured-sales@macnica.co.jp

*Company names and product names mentioned in this text are trademarks or registered trademarks of Macnica and each company.
*The information published in the news release (including product price, specifications, etc.) is current as of the date of announcement. Please note that the information may be subject to change without prior notice.