QommonsAI Equips Private Knowledge with a 'New Hierarchical Structure,' Fundamentally Changing AI Response Accuracy by Systematizing Internal Documents with Up to 5 Folder Levels and 3-Tier Role Permissions

Polimill Co., Ltd. (Headquarters: Minato-ku, Tokyo; Representative Directors: Ayame Ito / Nonoka Taniguchi), a top runner in administrative AX, announced today that it has begun offering a 'New Hierarchical Structure for Private Knowledge' in its generative AI for government, 'QommonsAI,' starting April 1, 2026.

Launch of "New Hierarchical Structure for Private Knowledge"

"Knowledge exists. But it's unusable" - A common challenge for local governments

In many local governments, documents are scattered across departments, and past knowledge remains buried and unused. Even with searches, necessary information cannot be found, and even if they want to refer to AI, they don't know "which data to organize and how."

Even if knowledge is accumulated, if it is not organized, AI cannot exert its power.

Responding to the fundamental problem of "inability to organize" with a system. That is the design philosophy of the new hierarchical structure.

"Intuitively organize" large amounts of knowledge

The new private knowledge is a "drive and folder hierarchical" knowledge management platform.

With a folder structure of up to 5 levels, internal government documents can be systematically organized by purpose, department, and task.

Three types of drives are provided. They can be used according to their purpose.

Screen image

"Rakuraku RAG" for internal folder structures, as is

One of the biggest features of the new hierarchical structure is the "Rakuraku RAG" function.

Simply upload documents accumulated on internal shared servers or file servers to private knowledge, maintaining the existing folder structure, and AI will immediately convert them into high-performance RAG. No need to rearrange or reclassify folders.

"Just bring the folders you have internally, as they are."

This eliminates the "data organization hassle" that has been a hurdle for RAG utilization, converting existing operational assets directly into AI knowledge. After uploading, it supports narrowing down the reference range by folder unit and providing links to source files. Documents that were dormant within the government will be reborn as information assets that support AI's response accuracy from day one.

"Who sees what" can be controlled at the system level by administrators

Simultaneously important with knowledge organization is access permission management.

The new hierarchical structure allows for fine-grained control over document access ranges by department and task, using 3 levels of user roles.

For example, HR and payroll-related documents can be limited to HR department members, while disaster prevention-related documents can be shared across the crisis management department and related departments—flexible operation that is "not too open, not too closed" is realized not by relying on operational rules, but as a system mechanism.

Protection against accidental operations is also standard

Even if documents are accidentally deleted, they can be stored for a certain period and restored using the trash function.

It also supports automatic determination of restoration destinations based on the parent folder's status, preventing document loss due to erroneous operations.

Furthermore, 100GB of free RAG storage is provided as standard for basic local governments, and 300GB for broader regional governments.

"AI that can use knowledge" is available from day one without additional costs.

Supported file formats

PDF / Word (docx) / Excel (xlsx) / PowerPoint (pptx) / Text (txt) / CSV

Character codes: Supports UTF-8, Shift_JIS, CP932

Organized knowledge changes AI's answers

By organizing knowledge in a hierarchical structure, the accuracy of AI's search and reference improves.

Batch selection by folder unit - Narrows down the reference range during chat, achieving noise-free answers

Link display of source files - Clarifies the source of information along with the answer, supporting evidence-based decisions

Cross-drive search - Searches files and folders across accessible ranges

Knowing "which materials are the basis for the answer." That's why AI's answers can be used for business decisions.

QommonsAI's private knowledge evolves from "a place to accumulate" to "a knowledge base that drives operations."

About QommonsAI

QommonsAI, a generative AI for government developed, provided, and operated by Polimill Co., Ltd., supports the resolution of local government challenges based on evidence from tens of millions of data points, including domestic and international laws, policies, papers, and local government case studies.

As of March 2026, it is used by over 700 local governments nationwide and approximately 250,000 people. It is utilized in a wide range of tasks such as parliamentary affairs, policy formulation, resident support, and public relations.

It is equipped with 12 carefully selected models from four major companies: OpenAI, Anthropic, Google, and Preferred Networks, providing an environment where the optimal AI can be selected according to the purpose.

It is provided directly to all local governments without intermediaries, and since its service release, on-site implementation training and introductory/intermediate training have been provided free of charge to all adopting local governments.

Furthermore, an agile development system is in place to reflect requests from local government sites into the service in as little as a few days, continuously delivering "currently needed functions" to the administration at overwhelming speed.

Official website: https://info.qommons.ai/

Company Profile

Company Name: Polimill Co., Ltd.

Location: Minato-ku, Tokyo

Representatives: Representative Directors Ayame Ito / Nonoka Taniguchi

Business Activities: Planning, development, and operation of generative AI for government "QommonsAI," digital democracy platform "Surfvote," etc.

Corporate website: https://polimill.jp/