Significantly Expanded "AI Agent Group for Business Planning Support" to Aid Business Planning

Stockmark Inc. (Headquarters: Minato-ku, Tokyo; CEO: Tatsu Hayashi; hereinafter referred to as "our company"), a provider of proprietary development of domestic generative AI platforms and generative AI services for businesses, is pleased to announce the development and implementation of two new agents (Technology Portfolio / Business Opportunity Evaluation) and a "Feedback Memory Function" for the Market Size Estimation agent, significantly expanding the "AI Agent Group for Business Planning Support" which enhances R&D and new business planning operations.

This expansion of the agent group enables R&D and new business divisions in the manufacturing industry to explore "business opportunities originating from technology" with overwhelming speed and high reproducibility, a task previously reliant on the experience of specialized personnel or external consulting firms. By having AI structure vast amounts of technical information and market feedback, and deriving "winning scenarios" based on the company's strengths, it powerfully supports internal consensus building and swift management decisions.

Background of Development

Our company has been supporting the advancement of the engineering chain, from R&D to quality assurance in the manufacturing industry, through our platform for autonomous AI operations, "SAT Agent Cockpit."

- Research and Development: Enhancing discovery of unknown needs and exploration of technology applications - Design: Shortening development lead times by structuring past design knowledge - Manufacturing: Inheriting expertise of skilled workers and optimizing processes - Quality Assurance: Automating early detection of market defects and risk management

Among these, how to connect a company's existing technologies with new market challenges and commercialize them early is a critical issue for many companies. However, in many cases, the exploration of applications and creation of new themes involve individual efforts such as hypothesis formulation based on the experience of engineers and planners, public information research, patent analysis, market size estimation, and preparation of internal presentation materials, leading to issues of personalization and fragmentation in the review process.

Furthermore, many on-site teams are facing new challenges when using general-purpose AI tools like ChatGPT in their actual work.

- The Wall of Generalities: Reports generated are textbook-like, and it's difficult to see the "winning strategy" for leveraging our company's technology. - The Disposable Wall: Even if you instruct the AI once with "Calculate with these preconditions," it resets the next time you use it, requiring the same correction instructions to be repeated each time. - Data Fragmentation: Patent information and daily updated market challenge data held by the company are not effectively connected to actual business planning.

This newly developed group of agents fuses our company's strengths in "knowledge graph technology," proprietary image and text analysis technology (VLM), and Deep Research (a mechanism where AI repeatedly verifies itself). This enables a seamless process from "building the foundation for analysis" in R&D and new business planning, to "deriving hypotheses for our company's winning strategy," and "a learning cycle that adapts to our organization with use (improving future outputs)."

Newly Developed Agents and New Functions

1: Technology Portfolio Agent: Visually maps technology and market challenges for instant opportunity identification

This agent automatically generates a map (matrix) of the intersection between "market challenges" and "technological approaches" for a specific theme, based on patent and technology information.

In traditional technology analysis, while it was possible to view the number of patents and distribution by technology classification, interpreting which "market challenges they correspond to" or "which areas are promising for business opportunities" required expert interpretation.

By structuring technology information and linking it to market challenges, this agent allows for an at-a-glance overview of "which technologies are concentrated for which challenges," "where competition is concentrated," and "where unmet needs (white space) exist," without relying on specialized knowledge.

- Functionality: Generates a matrix combining market challenges and technological approaches, visualizing the distribution of patent information. It can also overlay technology information structured by "SAT" onto the patent distribution, allowing for seamless connection to detailed analysis by other AI agents starting from areas of interest.

- Value: Instantly creates a map that shows "where to focus," enabling the entire team to specifically discuss "why we should delve deeper here" and "how our company's technology can be utilized." It significantly improves the quality of initial hypotheses for research theme selection and promotes swift decision-making across departments.

[DEMO VIDEO] https://www.youtube.com/watch?v=NFtOQW4JMKY

2: Business Opportunity Evaluation Agent: Deep Research leading to hypotheses for our company's winning strategy

This is a workflow-type agent that supports research target organization, issue design, report generation, and next action proposals by AI, simply by inputting the theme to be researched and company information.

As a core function for "technology-driven business opportunity exploration" envisioned by our company, this agent collaborates with the Technology Portfolio Agent, Market Size Estimation Agent, and Forced Idea Generation Agent to support the entire process from theme exploration to business viability assessment and idea concretization.

- Functionality: After the user confirms and edits the research scope presented by the AI based on the input theme and company information, the AI autonomously researches market/customer/competitor/technology trends. It evaluates "unique entry possibilities based on the company's existing technologies and focus areas," rather than general market overviews, and finally presents "winning strategy hypotheses" and "next steps."

- Value: Allows for the organization of "Is there really room for entry?" "Which customer challenges can it be applied to?" and "Which issues should be further verified?" within a single workflow. It efficiently builds business hypotheses directly linked to internal discussions and investment decisions.

[DEMO VIDEO] https://www.youtube.com/watch?v=gJh5ADhv7qY

3: Feedback Memory Function: AI learns user "preferences" and reflects them in future outputs

In the Market Size Estimation agent, this function allows the AI to remember and accumulate user-made corrections and feedback, automatically reflecting them in subsequent outputs.

Market size estimation results can vary significantly depending on how the target market is segmented, the preconditions set, the unit prices or adoption rates used, and the scope of regions or applications. In conventional general-purpose AI tools, preconditions and calculation policies once pointed out by the user were not sufficiently reflected in subsequent uses, requiring the same corrections and feedback to be made each time.

This function is based on the "Human in the loop" philosophy, where humans incorporate their judgment and preferences into the process to make the AI smarter, rather than relying solely on AI. It is a mechanism for continuously reflecting user corrections as organizational knowledge, rather than just a one-time fix.

In the future, in addition to market size estimation, deployment to other agents such as Business Opportunity Evaluation and Technology Portfolio Analysis is planned.

- Functionality: The AI learns feedback provided on the Market Size Estimation agent's output, such as "Estimate unit price around this much for this market" or "Exclude this region," as "reusable rules." These rules are not only reflected in future outputs but can also be input, updated, and deleted by the user at any time.

- Value: As the analysis policies specific to each company and the judgment criteria of each person in charge are accumulated in the AI, the agent grows to become more suitable for practical use the more it is used. It reduces the unproductive time spent correcting the same preconditions each time and allows for the efficient creation of analysis documents that align with the required granularity and logical structure within the company.

[DEMO VIDEO] https://www.youtube.com/watch?v=kK1mmero6KY

Next-Generation Business Planning Process Realized by the Agent Group

With this expansion, our company's AI agent group for business planning support has evolved into an ecosystem that supports the three steps of technology-driven business creation: "1 Exploration," "2 Evaluation," and "3 Learning."

- Exploration: Use the "Technology Portfolio" to identify untapped areas to focus on, from patents and market challenges. - Evaluation: Combine with our company's strengths using "Business Opportunity Evaluation" to calculate "winning scenarios" and "market size." - Learning: Accumulate user and company "preferences (planning know-how)" through "Feedback Memory."

This allows R&D and new business divisions, which were previously overwhelmed by individual data investigations, to proceed with hypothesis building, quantitative verification, and idea concretization for commercialization as a unified process.

Future Development: Decision Support Integrated with Company-Specific Data

The agents and functions implemented this time will be integrated with our company's autonomous AI operation platform, "SAT Agent Cockpit." By incorporating company-specific technical data, R&D information, patent information, customer challenge information, and past planning materials, we plan to support more advanced customization.

In particular, for R&D and new business divisions in the manufacturing industry, structuring each company's technological assets and tacit knowledge into a state that AI agents can utilize will enable the advancement of the business creation process itself, beyond mere efficiency improvements in research.

Furthermore, based on these agents, our Biz members and engineers can provide services in a format where they tune the system and deliver analysis reports. Results such as completing and delivering market analysis projects in one month, which typically take three months, have been achieved multiple times, offering significantly faster and higher-quality deliverables compared to normal consulting projects.

Our company will continue to support Japanese manufacturing and R&D companies in accelerating the speed of "application exploration" and rebuilding global competitiveness through the fusion of generative AI and proprietary technologies.

About SAT Agent Cockpit

SAT Agent Cockpit is a platform that enables the creation and operation of autonomous agents by converting a company's unique "tacit knowledge" and "complex unstructured data" into immediately executable assets (Agent-Ready).

With the latest Vision Language Model (VLM) technology, it structures not only text but also visual information such as drawings, specifications, and charts. Furthermore, by incorporating a process for formalizing the know-how of veteran employees, it achieves company-wide "AI BPR (Business Process Re-engineering)" that goes beyond simple operational efficiency.

- Official Website: https://sat.stockmark.co.jp/

About Stockmark's Solutions

AI utilization is an indispensable element for maintaining competitiveness. However, many companies face challenges such as "data not being properly organized," "lack of adoption at the operational level," and "failure to lead to concrete results."

Our company uses proprietary natural language processing technology to structure complex knowledge, including not only text but also drawings, specifications, and past decision logic, into a form that AI can utilize. This promotes "AI BPR (Business Process Re-engineering)" that redesigns the operational process itself, enabling people to focus on "value creation" and "skill refinement," which they should originally concentrate on, going beyond mere efficiency improvements.

By having AI autonomously handle "simple tasks that create stagnation" and shifting people to "high-value-added work," we aim to create a state where people can "genuinely enjoy their work" and boost the competitiveness of Japanese companies.

- Stockmark Solutions: https://stockmark.co.jp/solution/

About Stockmark Inc.

Stockmark Inc., with its mission to "reinvent the mechanism of value creation," supports the transformation of many companies by leveraging cutting-edge generative AI technology.

We operate "Aconnect," an AI agent for the manufacturing industry, and "SAT," which structures all types of data into company assets. Furthermore, we support the development of company-specific generative AI and the construction of proprietary systems.

Company Name: Stockmark Inc. Location: LIFORK MINAMI AOYAMA S209, 1-12-3 Minami-Aoyama, Minato-ku, Tokyo Established: November 15, 2016 Representative: Tatsu Hayashi, Representative Director and CEO Business Content: Development and operation of services that support corporate knowledge management and the application of generative AI in business, utilizing cutting-edge generative AI technology. URL: https://stockmark.co.jp/