KumaKumaAI Launches 'Note Function' for AI-Driven Knowledge Extraction and Organization — Moving Beyond Search-Only RAG to an Enterprise AI Platform That Gets Smarter with Use

Knowledge Notes

Grizzlarity Inc. (Headquarters: Chiyoda-ku, Tokyo; CEO: Tomoya Fujita) has launched the 'Note Function' on its industry-specific multi-agent AI platform 'KumaKumaAI' starting July 2026. This function enables AI to automatically extract key insights from internal documents and conversations, storing and organizing them as editable notes.

While the adoption of AI-powered document search systems (RAG) is growing, search results are limited to machine-split document fragments. Furthermore, vectorized data cannot be directly edited, making it difficult to address scattered or outdated information. With this new function, knowledge becomes continuously organized through AI-driven selection, integration, and classification. By storing insights as human- and AI-readable, editable text, KumaKumaAI enables both ongoing knowledge management and improved AI response accuracy.

Background and Challenges

In enterprise knowledge utilization, environments where documents can be uploaded and searched are becoming common. However, search results return only mechanically fragmented document snippets, and outdated content cannot be directly corrected by humans. Additionally, the same facts are scattered across multiple documents, with no clear way to identify the latest version, leading to ever-growing information without proper organization.

This results in a structural limitation where 'search is possible, but knowledge keeps growing without being organized,' preventing further improvement in AI response quality.

Overview of the Note Function

The 'Note Function' adds a new layer of knowledge—'AI-extracted, human-editable insights'—on top of existing document search.

Key features include the following five:

AI-Powered Automatic Extraction: The AI automatically extracts only high-value insights from documents or conversations for future reference. To prevent note overproduction, extraction is limited (e.g., up to 10 notes per document), filtering out transient information. Extraction policies can be defined in natural language for each project.

Automatic Deduplication and Integration: When new information matches an existing note's topic, the system prioritizes integration into the existing note rather than creating a duplicate. This maintains a 'single source of truth' and prevents confusion over which version is current.

Human-First Collaborative Editing: The master version of each note is a human-readable and editable Markdown text. Users can directly edit notes on-screen, and changes are immediately reflected in future searches and AI responses. When AI updates conflict with human edits, the system switches to a 'merge proposal' mode—highlighting differences and awaiting human approval instead of overwriting. All changes, whether by AI or humans, are tracked in version history for full traceability.

Verifiability via Source Links: Each note retains source links to the original documents, allowing users to trace AI-generated content back to its original context at any time. Notes can also be linked to each other, enabling navigation across related knowledge.

Automated Curation: The system automatically runs periodic tasks such as proposing note integrations, detecting contradictory content, and flagging notes whose source documents are missing. Folder structures are not fixed templates; instead, AI automatically builds and reorganizes them based on accumulated content, suggesting folder splits and reorganizations. The system respects human decisions—for example, integration proposals rejected once will not be repeated.

AI Agent Utilization of Notes

KumaKumaAI agents first consult the organized note layer when generating responses, only delving into original documents when necessary. Beyond search, agents can autonomously create, update, and reorganize notes and folders based on chat instructions, enabling real-time knowledge capture from conversations. Users can specify which notes to reference or choose the search scope (notes, original documents, or both). As knowledge accumulates and becomes better organized, response accuracy and speed improve.

AI agents using the note search tool

Use Cases

[Customer Support]

AI automatically generates troubleshooting and FAQ notes from past support tickets and incident reports. When similar issues recur, agents instantly respond by referencing the organized notes. If procedures change, staff can directly update the notes, and all future responses will reflect the latest information.

[Sales and Customer Management]

Key customer details, histories, and unique agreements are extracted from meeting minutes and client communications and stored as notes. During personnel transitions, new staff can quickly understand a client's status from up-to-date, organized notes—without needing to review scattered meeting records.

Service Overview

Launch Date: July 2026

Deployment Options (choose one of three):

SaaS Version: Low-cost deployment via a shared cloud environment

Self-Hosted Version: Deployed within the customer’s AWS environment (VPC), ensuring data remains within the company’s infrastructure

Domestic GPU Base Version: Utilizes Sakura Internet’s high-performance VRT, completing all processing—including LLM inference—within domestic data centers. Operated on a fixed-fee basis

Pricing: Available upon inquiry (custom quotes)

How to Apply: Contact via https://grizzlarity.com/contact

Exhibition at 'AI-Driven Development Conference 2026 Summer'

We will exhibit at the 'AI-Driven Development Conference 2026 Summer' on July 30 (Thu) and 31 (Fri), 2026. At our booth, visitors can experience live demos of KumaKumaAI, including the Note Function. On July 30, our CEO Tomoya Fujita will deliver a presentation on business automation using AI agents.

Future Outlook

Going forward, we will enhance classification systems based on note accumulation patterns and expand integrations with external tools. Our goal is to realize a knowledge infrastructure where organizational knowledge becomes better organized with use—evolving into a valuable asset beyond mere searchability.

Grizzlarity Inc.

Address: Yamato Building 405, 1-6-16 Kanda-Izumicho, Chiyoda-ku, Tokyo 101-0024, Japan CEO: Tomoya Fujita, Representative Director Business: Development of enterprise multi-agent AI systems, provision of deep research AI agents, and construction of AI research support platforms URL: https://grizzlarity.com/ Contact: https://grizzlarity.com/contact