Knowledge creation support system using AI specifically for internal organizations

Abstract

We provide a knowledge creation support system that enables organizational learning to promote company-wide growth by learning and reproducing a company's culture, rules, and decision-making criteria. [Solution] The knowledge creation support system comprises an AI solution 1 as an expert knowledge learning and reproduction engine that replicates the thought processes of experts within the organization, learns and utilizes organization-specific knowledge, and has the function of continuously learning and improving; an integrated platform 2 that links the data accumulated in AI solution 1 with existing systems, integrates data sources, and performs access control and authentication management; and a data storage 3 that manages large-scale data accumulated through data storage. The AI ​​breeder 4 implements the planning of digital transformation strategies and creation of AI utilization roadmaps before the introduction of the system to the organization, system construction and data integration support during system introduction and operation, and organizational transformation support.

Description

Technical Field

[0001] The present invention relates to a knowledge creation support system for an entire organization by a dedicated AI within an organization that integrates an AI agent with self-executing power, an AI assistant with information support power, and an AI companion with continuous accompanying power, and is not just a tool but an "organization member" that grasps the company's unique knowledge and judgment criteria, such as an AI officer, AI department head, AI section head, and professionals in a company, enabling 24 / 7 / 365 business support.

Background Art

[0002] So far, technologies have been proposed to protect information used in generative AI securely while protecting company-specific and industry-specific information from leakage (Patent Documents 1 and 2). However, these technologies are not about the flow of the overall organization structure but about protecting information security in the construction industry and educational settings.

[0003] Also, general-purpose AI such as general-purpose chat services has been popular. For example, ChatGPT, Gemini, Claude, etc. are widely used. However, these general-purpose chat services only have general knowledge and cannot learn a company's unique rules, know-how, and judgment criteria. They cannot understand the in-house specialized knowledge and tacit knowledge accumulated daily, so they cannot provide answers in line with the company's specific context.

[0004] General-purpose chat services are fixed models and cannot continuously learn and improve according to the business of each company. Even if there are changes in specific business processes or judgment criteria, the AI does not evolve accordingly. A company cannot grasp and manage who is using it and what effects it has had. There is a lack of a mechanism to quantitatively evaluate the introduction effect of AI tools and measure the investment return.

[0005] Furthermore, with general-purpose chat services, input content may be used as training data, posing a risk of confidential information being leaked externally. When handling important internal information or customer data, security concerns limit its scope of use. In addition, because the user's position and authority cannot be understood, it is impossible to provide appropriate level of answers or decision-making support. The level and content of answers required differ between newcomers and veterans, and between field staff and managers. Moreover, only general answers can be provided, and judgments based on deep industry-specific expertise and practical experience are not possible. It is difficult to provide answers that fully understand the industry-specific terminology, regulations, and practices of industries such as manufacturing, finance, and healthcare.

[0006] In other words, general-purpose AI possesses "global knowledge," but it doesn't know about individual companies. No matter how advanced a general-purpose AI may be, it cannot understand the crucial company-specific knowledge such as the history, culture, unwritten rules, past successes and failures, organizational structure, and business processes of individual companies. Without this information, truly valuable decision-making support is impossible. To overcome this fundamental limitation, there was a need for specialized AI solutions tailored to specific companies (organizations).

[0007] In 2022, ChatGPT emerged, and it became recognized that "people who can master LLMs (Large-Scale Language Models) will have an advantage." On the other hand, humans (real-world human beings) were feeling the limits of their own brains ("thinking," "remembering"), and were in a situation where they could experience brain fog due to overwork.

[0008] Given this situation, a digital twin AI was developed that complements the human brain, specifically one equipped with an external engine for "thinking" and an external storage for "remembering" (Waseda University smart SE IoT / AI course (2023)). Thus, there was a demand for a business system that would allow individuals to develop and utilize their own AI in areas where they felt human limitations, and to grow together with their own AI. [Prior art documents] [Patent Documents]

[0009] [Patent Document 1] Japanese Patent Publication No. 2011-221584 [Patent Document 2] Japanese Patent Publication No. 2009-093538 [Overview of the Initiative] [Problems that the invention aims to solve]

[0010] The present invention aims to provide a system that learns and reproduces the culture, rules, and decision-making criteria of a company, thereby nearly completely replicating the organization to function as a "true in-house AI." It operates in a dedicated in-house environment, has no risk of information leakage, and is in a nearly completely secure environment where confidential information can be discussed with confidence. The system promotes organizational learning so that individual learning and experience contribute to the overall knowledge improvement of the organization, thereby achieving company-wide growth. [Means for solving the problem]

[0011] As a result of diligent research, the present inventors have achieved the above objective by providing the following invention.

[0012] 1. An AI solution as a specialized knowledge learning and reproduction engine that replicates the thought processes of experts within an organization, learns and utilizes organization-specific knowledge, and has the functionality to continuously learn and improve. An integrated platform that links the data accumulated by the aforementioned AI solution with existing systems, integrates data sources, and performs access control and authentication management, A system comprising a data storage that manages large-scale data accumulated from the aforementioned data and has an automatic backup function, A knowledge creation support system implemented by AI Breeder, which includes formulating a digital transformation strategy and creating an AI utilization roadmap before the introduction of the system to the organization, supporting the construction of the system and data integration during the introduction and operation of the system, and supporting the transformation of the organization.

[0013] 2. The knowledge creation support system described in 1. above, wherein the organization is operated through collaboration between real members of the organization who possess the aforementioned expertise and corresponding digital members of the generated AI.

[0014] 3. The AI ​​solution comprises a program that causes a computer to function in order to generate the digital subject in cyberspace using text data relating to information about the real-world members. The program is a knowledge creation support system as described in 2., comprising: a knowledge unit that includes multiple RAG (Retrieval-Augmented Generation) databases for the digital member to refer to when generating answers to user questions; a personality unit that represents information on the personality and characteristics of the digital member; a habit unit that performs actions to be generated according to user instructions; and an engine unit consisting of an LLM.

[0015] 4. The knowledge creation support system described in 1. above, which includes an AI breeder certification system within the organization aimed at developing personnel capable of developing AI solutions specialized for various tasks, with multiple certification levels corresponding to skill levels and roles, and providing support for the introduction of the AI ​​solutions to the organization and / or external organizations other than the organization that do not have AI breeders. [Effects of the Invention]

[0016] According to the present invention, the culture, rules, and decision-making criteria of a single organization (a single company) are learned and reproduced exactly as they are, and the organization is almost completely replicated to function as a "true in-organization AI". It operates in a dedicated environment within the organization, has no risk of information leakage, and is in an almost completely secure environment where confidential information can be discussed with peace of mind. Individual learning and experience contribute to the improvement of knowledge throughout the organization, and organizational learning can be promoted to achieve company-wide growth. [Brief explanation of the drawing]

[0017] [Figure 1] As an example of the configuration of a data processing system by the system of the present invention, it is a conceptual diagram showing a multi-step thinking engine (LLM) that operates multiple times on an unrestricted RAG database. [Figure 2] It is a conceptual diagram showing the relationship between an agent and a database in a conventional AI tool. [Figure 3] It is a conceptual diagram showing an example of the relationship between an agent and a database in the system of the present invention. [Figure 4] It is a conceptual diagram showing an overall view of an example of a program that constitutes an AI solution in the system of the present invention. [Figure 5] It is a conceptual diagram showing an example of user utilization in the system of the present invention. [Figure 6] It is a conceptual diagram showing a comprehensive system by the implementation of the system of the present invention.

Embodiments for Carrying Out the Invention

[0018] The knowledge creation support system of the present invention, as described above, includes an AI solution (AI digital twin, expert knowledge learning and reproduction engine), an integrated platform, and data storage, and is implemented by an AI breeder.

[0019] Also, in the system of a preferred embodiment of the present invention, by providing the system prompt and RAG data to the LLM, it is possible to reflect human personality and knowledge. By having the habit (note) of "organizing one's mind" with AI and repeatedly providing the values and knowledge of real (real) humans to the AI, it becomes possible for "real humans themselves to grow" by interacting with the AI purposefully.

[0020] The system of this invention implements a knowledge transformation process called the "SECI model" (proposed by management scholar Ikujiro Nonaka) and supports knowledge creation throughout the organization. In knowledge creation theory, "tacit knowledge" refers to knowledge such as individual experiences, know-how, and intuition that are difficult to express in words or numbers, and it plays an important role. This model is a framework that promotes the creation and expansion of organizational knowledge by converting tacit knowledge into explicit knowledge (knowledge that has been verbalized and quantified) and circulating the two.

[0021] A system has been proposed that will continuously grow by learning the unique information of a company and collecting data together with the organization. It would gather the company's proprietary knowledge, industry-specific terminology, and past decision-making examples, and then act as a model or partner to support the work and provide advice.

[0022] In the system of the present invention, a warehouse-like structure can be used to store data and generate AI, and a system can be used to take in data from an external source and utilize the open-source Elasticsearch mechanism.

[0023] (AI Breeder) The AI ​​Breeder formulates digital transformation strategies and creates AI utilization roadmaps before system implementation within an organization. Furthermore, during system implementation and operation, it builds systems, provides data integration support, and assists in organizational transformation. The purpose of establishing the AI ​​Breeder is to address issues such as information leakage that can occur when outsourcing AI system work to external parties; instead, it ensures that information is configured internally without being shared externally. The AI ​​Breeder plays a crucial role in the comprehensive system implementation of the present invention, as shown in Figure 6.

[0024] To apply the system according to the present invention, an application as an AI solution (for example, "A:nmoku-ch:I TMWhen introducing and operating the commercially available product (manufactured by Keepdata Inc.), an AI breeder is implemented within the company. When using this AI solution within a company, it is assumed that personal information of a certain "person" within that "company"—such as date of birth, age, place of origin, height, weight, etc.—cannot be released (disclosed) outside the company. Therefore, the AI ​​breeder is created and prepared within the company, and the AI ​​breeder is configured with the company's internal information.

[0025] The AI ​​breeder will teach employees about the "SECI model" and how to incorporate company data into AI solutions. Having the SECI model enables a cycle of visualization of AI solutions, systematic knowledge sharing, and continuous learning. This allows the present invention to provide an integrated platform and a complete knowledge creation support system, including solutions. The system of the present invention, in particular, can provide a framework implemented as a knowledge creation and transmission model through collaboration between humans and AI by implementing the HAC-SECI model (Human-AI-Collaboration SECI Model), which is an evolution of the SECI model (knowledge transformation through joint creation, externalization, linking, and internalization). While the conventional SECI model was limited to knowledge transmission between humans, the HAC-SECI model, through collaboration with AI, enables 24 / 7 knowledge access, one-to-many digital twin creation, and inheritance that transcends space and time. The implementation of the HAC-SECI model realizes a collaborative knowledge transfer process between humans and AI, completely overcoming the limitations of traditional human-to-human knowledge transfer, such as time, place, and individual differences, through AI technology, and enabling continuous knowledge creation across the entire organization. Specifically, it enables a four-stage knowledge creation process: extraction of tacit knowledge through natural dialogue with AI using an interview AI system, experience sharing platform, and tacit knowledge visualization engine (collaboration); formalization of structured knowledge with AI support using a thought process structuring AI, knowledge framework automatic generation, and judgment criterion documentation system (externalization); knowledge integration and optimal solution extraction by AI using a knowledge graph automatic construction, pattern recognition, analysis AI, and best practice extraction engine (combination); and personalized adaptive learning (internalization) through AI coaching based on an adaptive learning AI, growth support engine, and continuous feedback system. A preferred embodiment of the system of the present invention is the aforementioned "A:nmoku-ch:I" which implements such a HAC-SECI model. TM By adopting this approach, continuous knowledge creation can be more effectively achieved throughout the organization.

[0026] In this invention, in order to effectively utilize in-house AI solutions without distributing them outside the organization, a system for a novel business model using AI breeders can be provided. To further utilize the AI ​​solutions related to the system of this invention, it is desirable to increase the number of AI breeders. (AI Breeder Certification System) Furthermore, in the system of the present invention, by establishing an AI breeder certification system, certified AI breeders capable of handling applications as AI solutions, such as the aforementioned "A:nmoku-ch:I TM AI breeder utilizing " (A:nmoku-ch:I TM As an AI promoter within the company, the AI ​​breeder leads the use of AI solutions in each department, designs and implements automation and optimization of daily operations to improve business efficiency by utilizing confidential company data without sharing it outside the company, and is responsible for formalizing tacit knowledge and providing training data for AI. The AI ​​Breeder Certification System is a system where employees with specialized knowledge within the company are certified by A:nmoku-ch:I TMThe aim is to cultivate human resources capable of developing AI solutions specialized for various tasks by utilizing these resources. Furthermore, it aims to improve the autonomy of certified companies, optimize licensing costs, and further enhance A:nmoku-ch:I TM This aims to expand the ecosystem and improve quality. The AI ​​breeder certification system offers multiple certification levels tailored to skill levels and roles, with different learning content, certification requirements, and benefits at each level, allowing for gradual skill development. From beginners to experts, a growth path is provided that matches each individual's skills and goals. At each level, users build upon the skills acquired in the previous level, gaining more advanced knowledge and practical abilities. Certification requirements become progressively stricter, but the benefits also increase accordingly. This helps maintain motivation for continuous learning and skill development. Furthermore, certified AI breeders can also be utilized by external organizations (companies, etc.) that do not have their own AI breeders, enabling them to improve operational efficiency by utilizing confidential internal data within those external organizations without disclosing it outside the company. In this way, certified AI breeders can undertake implementation support services not only within their own companies but also externally. In addition, they can gain an advantage in job changes and promotions by increasing their personal market value and establishing their expertise.

[0027] (AI Solutions) The AI ​​solution in the present invention is, so to speak, an AI that "works" like a "human." This AI solution is not merely an AI tool, but a next-generation AI solution that functions as "a member of the organization." Unlike general-purpose AI, it can learn the culture, rules, and decision-making criteria of individual organizations (companies, companies) and provide support tailored to that organization.

[0028] In a system according to a preferred embodiment of the present invention, an AI solution integrating conventional AI functions is employed, specifically integrating three functions: an AI agent function, an AI assistant function, and an AI companion function, which operate as the user's "company-specific AI executive" (AI digital twin).

[0029] Here, AI digital twin is a technology that further enhances the "digital twin," which uses AI (artificial intelligence) to reproduce the real world in a digital space with high accuracy and in real time. While conventional digital twins reproduce and simulate the real world, AI digital twin automates data analysis and learning, enabling faster and more accurate predictions, improvement suggestions, and decision-making, thus functioning as a "self-evolving digital twin."

[0030] The AI ​​agent function enables autonomous task execution and business process automation. Based on predefined rules and learned workflows, it performs complex tasks without human intervention. The AI ​​assistant function provides support for information retrieval, organization, and analysis. It instantly extracts necessary information from vast amounts of internal company data and presents it in an easy-to-understand format. The AI ​​companion function provides continuous relationship and long-term support. It learns the user's work characteristics and preferences, and provides optimal support through the building of a long-term relationship.

[0031] In the system according to this embodiment, the AI ​​solution organically integrates the three functions described above, enabling comprehensive business support that was not possible with conventional AI tools. This goes beyond mere information provision and task automation, fundamentally transforming an organization's knowledge creation and decision-making processes.

[0032] The AI ​​solution in the present invention is a customizable conversational AI solution, fundamentally different from conventional AI, being a "continuous learning AI." In other words, this AI solution is not just a chatbot, but an AI that grows together with the company. By learning the thought patterns, values, judgment criteria, and expertise of individuals, veterans, and organizations, and by continuously learning through user feedback, it complements the human tasks of "remembering," "thinking," and "searching," and provides "insights."

[0033] The three main features of the AI ​​solution according to the present invention are (1) persona / character setting (anthropomorphism), (2) unlimited RAG database, and (3) multi-stage thinking engine (LLM). Each of these elements is explained below.

[0034] (1) Persona / Character Setting (Personification): By specifying and setting age, gender, background, characteristic way of thinking, and data, the system generates responses as if it were that person. This is effective for handing over tasks and for tasks that are highly dependent on specific individuals. As a result, even when veterans retire, their know-how is accumulated as a corporate information asset, and new knowledge can be added and updated through feedback.

[0035] (2) Unlimited RAG database: Multiple databases that serve as the source for AI response generation can be set up in parallel without limit. By connecting with enterprise systems and databases, cross-data utilization of AI is possible (see Figure 1).

[0036] (3) Multi-Stage Thinking Engine (LLM): Unlike typical AI, this system deeply infers and understands the user's intent, improving the accuracy of its responses. For a single user query (input), the LLM is activated multiple times to perform actions such as "intent inference" and "metadata extraction" for information retrieval, allowing it to search for more relevant information from sources such as RAG DB and file storage.

[0037] The AI ​​solution in the present invention has a crucial difference from other existing AI tools. Other AI tools, such as Copilot, My GPTs, and Gemini(gems) (registered trademark, hereinafter the same), are used as utility tools or single-theme Q&A bots. As shown in Figure 2, this has a one-to-one relationship between the agent and the database, which has limitations in enterprise settings. Because all data is stored in a single database in a "1 agent = 1 DB" relationship, there are limitations in search accuracy and management. Furthermore, there are limitations such as the lack of a feedback mechanism so answers do not improve, the inability to perform log analysis, and the difficulty in maintaining consistent character settings and "individuality" over the long term. The use cases of these AI tools are suitable for simple Q&A, FAQs, and routine tasks.

[0038] In contrast to conventional AI tools, the AI ​​solution in the present invention allows for the construction of an organizational knowledge management infrastructure, as shown in Figure 3. Specifically, the main features of the AI ​​solution according to the present invention include the ability to create an unlimited number of agents and databases within a single system, the ability for agents to freely connect to databases within the system, the ability to perform time-series analysis and conditional searches, the ability to switch to the optimal LLM, the growth potential of being able to set a feedback database for each agent, and the ability to perform log analysis of which RAGs were hit. Furthermore, by combining character files, RAG databases, and personality settings as character settings, it becomes possible to consistently reproduce "statements that are characteristic of that person" and "judgments that are in line with that person's habits and values."

[0039] The system of the present invention is configured with a program for executing an AI solution. Figure 4 shows an overview of the program as a preferred embodiment of the present invention. As shown in Figure 4, this program comprises, as an important part, an agent unit including a knowledge unit, a character unit as a personality, and a habit unit. This information can be obtained by inputting it into the LLM, which serves as the engine.

[0040] The Knowledge Section is the database that the AI ​​references when generating answers. This is a RAG database, which combines a search engine and generative AI to address advanced information retrieval and response generation technology. By providing information that the AI ​​has not initially learned, it ensures that the answers are aligned with appropriate content.

[0041] The character section is where you input basic information about the "person" (real member), such as their personality, character, and characteristics, including how open-minded they are, how sincere they are (the five factors that make up the Big Five personality traits: 1. Openness to Experience, 2. Conscientiousness, 3. Extraversion, 4. Agreeableness, 5. Neuroticism), and their birthday. This section is where you input more detailed information that delves deeper into the person's character.

[0042] The habit section has a mechanism that performs actions when you input commands such as "Generate an image" or "Search for a code."

[0043] (Examples of how personas work) The system prompts in the program, as shown in Figure 4, define the AI's behavior, personality, and character. By incorporating not only the individual's dialogue history but also feedback from third parties and information from the HR system into the RAG DB, it gains deeper and more diverse knowledge about the individual's career, experience, and personality, becoming a digital twin of the user.

[0044] You can create your own digital twin by pre-entering information into the system prompt. Entering information into the system prompt should be done as shown in the example on the settings screen in Table 1.

[0045] [Table 1]

[0046] Furthermore, by storing feedback from third parties in RAG DB, a digital twin can be created from a more multifaceted perspective. For example, as shown in the example in Table 2, it can include not only the history of individual conversations with colleagues, superiors, team leaders, clients, client managers, client representatives, and subordinates, but also feedback from third parties and information from HR systems.

[0047] [Table 2]

[0048] (AI solution response output process) Here, we show an example of the output of the AI ​​solution. The UI (User Interface) and UX (User Experience) can be optimally built and provided to suit each use case.

[0049] When a user types "What topics were we discussing around this time last year?", the system accurately extracts past conversation logs using a summary of past conversations + RAG DB search (Table 3) and date search (Table 4), and generates a contextually appropriate response (Table 5).

[0050] [Table 3]

[0051] [Table 4]

[0052] [Table 5]

[0053] (Persona Mechanism: Multi-layered RAG Database) For an AI agent to become a truly effective business partner, it needs more than just information retrieval capabilities; it needs to understand the "individuality" of the user or organization—that is, their unique experiences, values, and how they are perceived by others—and behave in a way that is appropriate to that context.

[0054] Input: The following are examples of information that serve as the source of personas (RAG database (hereinafter also referred to as "DB")): For individuals: • What you possess: body, mind (brain), personality (Big 5), preferences • Experience: Career history, life log, events and incidents • Abilities: Technology, economic power, knowledge, uniqueness, authority • Internal environment: private space, lifestyle, family relationships • External environment: external profile, career, workplace relationships, circumstances Rules: philosophy, policies, routines, habits, preferences In the case of an organization: • What it contains: geography, people, ideas • Experience: History, successes and failures, events and incidents, achievements • Capabilities: Services, budget, resources, infrastructure, rights • Internal environment: Executive structure, culture / style, domestic relations, personnel information External environment: social responsibilities, position, diplomatic relations, circumstances • Rules: Business processes, laws and regulations, quality standards, constraints Here, there are static and dynamic settings. Furthermore, there are internal and externally influenced elements.

[0055] Output: Rules for AI behavior and growth Examples of behavioral rules (Practices) Thinking pattern: For complex questions, think logically step by step. Language use: Respond using appropriate language and format depending on the situation. Basic stance: Always adhere to the rules of conduct (ground rules). Rules for Continuous Growth Learning from dialogue: Continuously improve behavior and responses through user feedback. Following you: The AI ​​will continuously evolve in line with the changes in you, the original "you."

[0056] The system of this embodiment is further equipped with a feedback mechanism. For example, when a feedback button is pressed, the information is stored in a database (the communication holder in the knowledge section). As a result, when a question is entered in the future, the AI ​​grows by reflecting the previous data. In other words, a large database is accumulated, and the accuracy of the answers improves.

[0057] In conventional systems, the creation of AI agents and RAG infrastructure ends with the initial input information. In contrast, the system of the present invention allows the user to continuously add and adjust information to the initial setup while maintaining control, thereby improving accuracy.

[0058] (3-stage LLM processing) In a preferred embodiment of the present invention, the "three-stage LLM processing" technology enables the "automatic generation of appropriate information retrieval queries from ambiguous words." While general LLM and GPT with RAG generate answers in a single LLM call, the AI ​​solution of this embodiment significantly improves the accuracy of information retrieval by performing LLM processing three times for a given question. The three stages are Step 1: Advanced intent understanding, Step 2: Metadata extraction, and Step 3: Optimized integrated search.

[0059] In Step 1, LLM (Limited Licensing Management) is activated to infer and understand not only the superficial words but also the true underlying "intent" from the user's questions. For example, from a vague expression like "an urgent, important client matter from last month," it simultaneously understands the timing, importance, type, and urgency.

[0060] In Step 2, while understanding intent, LLM operates in parallel to extract specific conditions (metadata) such as dates and periods from the question, and prioritizes (biases) searches of RAG data and cloud storage. For example, for "contract documents from one year ago," the search is adjusted to prioritize files from one year ago.

[0061] In Step 3, based on the information obtained in Steps 1 and 2, LLM automatically generates the optimal search query to accurately identify and integrate the appropriate information and files from RAG data and cloud storage, and then displays the final results. This enables highly accurate search and response generation that would be impossible for a human to conceive.

[0062] This three-stage LLM process offers an innovative technological advantage, enabling implementation in a short period of time. Furthermore, the same logic can be applied to questions other than dates, allowing for appropriate searches based on intent. While traditional methods required 30 minutes of trial-and-error searching and inquiries with colleagues, the organizational effect is that every employee has a dedicated AI partner, enabling a work style where tacit knowledge is automatically shared, allowing even new employees to access the expertise of veterans, and automating the discovery of experts across departments.

[0063] (Data storage) The data storage in the present invention manages large-scale data accumulated from AI solutions and has an automatic backup function. It is not Hadoop-based, but employs an object storage (enterprise object storage) mechanism.

[0064] The core mechanism of object storage lies in managing data in units called "objects." Each object consists of "data itself," "metadata," and a "unique identifier," and is stored in a flat space without a hierarchical structure. This allows for the efficient storage and management of large amounts of unstructured data, and also offers excellent durability and scalability.

[0065] (Integrated platform) An example of an integrated platform in the system of the present invention is "AI.Doc.HUB" TM (Commercially available product, manufactured by Keepdata Inc.) etc. can be preferably adopted. This integrated platform can incorporate various types of internal data by utilizing generation AI-OCR, agents, and log collection tools. Complex data can be centrally managed by using object storage. Furthermore, it enables high-speed cross-search and full-text search, and visualizes the data. It provides a foundation for easily utilizing buried data, unique internal know-how, and person-dependent knowledge and information with generation AI, eliminating data silos within the organization.

[0066] Elasticsearch is a fast, scalable, distributed full-text search and analytics engine that indexes massive amounts of data in real time, enabling full-text search and complex analysis.

[0067] Furthermore, it can integrate with GitHub (a platform for sharing and managing source code) and connect with Salesforce (a cloud-based business platform that supports sales, marketing, and customer support). Connectors and MCPs can be used for this purpose. In this way, storage can be combined and provided as an ecosystem, each component can be used individually, and when combined into a set, it becomes even more powerful.

[0068] [Examples] Digital A (AI) The real person A, who is being coached, is a person who has worked as an engineer for a major company for over 28 years, involved in various projects. When the system in this example was used to ask Digital A, who had been inputted with information about Real A, "What are your favorite movies?", he answered "Back to the Future" and "Blade Runner". Both of these movies are in the fields of science fiction and technology, and they express the concepts of facts and time as entertainment while being conscious of the future and the past.

[0069] Based on the information about the real Mr. A, it is likely that he gave the above answer because his resume includes a lot of experience he has gained as an engineer over many years. Although the above questions were not asked to the real A beforehand, upon actually confirming with the real A, it was confirmed that they like both "Back to the Future" and "Blade Runner." Thus, it can be seen that the system of this embodiment can understand the thinking, personality, and background of the real person A, and provide information that has not been taught or inputted.

[0070] (Examples of user use) Figure 5 shows an example of how the system of the present invention can be used by a user. As shown in Figure 5, when a real person A (human) performs daily tasks while engaging in daily conversations with a digital AI A, the digital AI A accumulates characteristics of A, and the real person A also grows.

[0071] While preferred embodiments of the present invention have been described above, the present invention is not limited thereto.

[0072] The knowledge creation support system of this invention enables organizational reform, transforming organizations from traditional structures to AI-supported organizations. Specifically, everyone can access the expertise of veterans, democratizing knowledge; daily work is accumulated as organizational knowledge through continuous learning; and rapid talent development is achieved through personalized education by AI partners. Furthermore, an integrated knowledge base through cross-departmental collaboration enables company-wide optimization, significantly shortens the talent development period, improves knowledge transfer rates, increases the frequency of information sharing through inter-departmental collaboration, and fosters innovation by increasing new ideas through the new combination of existing knowledge. [Industrial applicability]

[0073] This invention has industrial applicability as a knowledge creation support system that learns and reproduces the culture, rules, and decision-making criteria of a single organization (a single company) as they are, and functions as a "true in-organization AI" by almost completely replicating the organization, operating in a dedicated environment within the organization, without the risk of information leakage, and in an almost completely secure environment where confidential information can be discussed with peace of mind, and which promotes organizational learning so that individual learning and experience contribute to the improvement of knowledge throughout the organization and realize company-wide growth. [Explanation of Symbols]

[0074] 1…AI Solutions (Digital Twin) 2…Integrated Platform 3…Data Storage 4…AI Breeder

Claims

1. A knowledge creation support system that generates digital twins of individuals with specialized knowledge within an organization, and supports the learning, utilization, and transfer of the organization's unique knowledge, It consists of a computer comprising at least a data processing unit, a knowledge unit, and a personality unit. The knowledge unit stores a vector database containing knowledge data including the subject's work flow, dialogue logs, and judgment criteria. The personality unit stores agent setting data that defines the personality characteristics and expertise of the subject, The aforementioned data processing unit An input process that accepts questions from the user in natural language, A query extension process that estimates the intent of a question from the input question using an intent estimation agent and generates a search query, Based on the generated search query, a RAG (Search Enhancement Generation) process is performed to execute a vector search on the vector database, and after correction through scoring and meta-search, relevant knowledge chunks are extracted. A response generation process that reflects the personality and knowledge of the subject by providing a system prompt based on the agent configuration data and the RAG data obtained by the RAG processing to an LLM (Large-Scale Language Model), A knowledge creation support system that performs this task.

2. The knowledge creation support system according to claim 1, wherein the data processing unit further obtains user feedback on the generated answers and stores the feedback as knowledge data in the vector database, thereby performing a learning process to continuously improve the accuracy of the answers of the digital twin.

Images