TIGEREYE Develops "Hierarchical State Intelligence (HSI)," a Proprietary AI Model That Predicts Future States and Actions, and Files a Patent Application for Its Core Technology

From AI that predicts “the next word” to AI that predicts “the next state and action.” Integrating state transitions, experiential learning, and hierarchical decision-making to advance autonomous AI agents and bring them into industry

TIGEREYE Inc. (Headquarters: Chuo-ku, Tokyo; CEO: Gaku Kamimura; hereinafter “TIGEREYE”) announces that it has developed “Hierarchical State Intelligence (HSI),” a proprietary AI model that predicts future changes in state and the next action to take based on the current situation, past experience, goals, and action history, and has filed a patent application for its core technology.

HSI is a proprietary AI architecture that places the relationship between “State,” “Action,” and “Next State” at the center of learning. It takes a different approach from large language models (LLMs), which have developed mainly by learning to predict the next word or sentence, and aims to support goal-aligned decision-making while predicting changes in the situation and the results of actions.

The model is built on three core technologies: “State Learning,” which learns the relationships among multiple pieces of state information; “State Transition Prediction,” which predicts actions and the changes in state that follow; and “Hierarchical Intelligence,” which evaluates the alignment between higher-level goals and individual actions. By also incorporating a mechanism that draws on past experience, it supports choosing actions based not only on short-term results but also on future impact and alignment with goals.

TIGEREYE is conducting basic verification of HSI’s fundamental learning and prediction functions, and will next advance verification of its performance, versatility, and practicality in real business environments.

This patent application protects, as intellectual property, the core technology of the AI model TIGEREYE is independently researching and developing, and strengthens the technology foundation for future business development.

Going forward, TIGEREYE aims to apply HSI to industrial fields such as AI agents, enterprise operations automation, healthcare, manufacturing, security, and robotics, and will put its research results into practical use in society through integration with its proprietary AI platform “NERVA” and adoption support from Forward Deployed Engineers (FDEs).

■ Background: “intelligence that decides,” what is needed next after generative AI

In recent years, the evolution of generative AI, centered on large language models, has enabled AI to demonstrate advanced capabilities in generating text, images, and programs, and in processing vast amounts of information.

In real society and business, however, it is not enough to generate information. What is required is to grasp constantly changing situations, predict the future based on past experience, and choose appropriate actions.

For example, when a company project falls behind schedule, analyzing and explaining the cause of the delay does not by itself solve the problem. It is necessary to take a comprehensive view of current progress, staffing, deadlines, past responses, and the goals of the organization as a whole, and to judge which action is most likely to improve the situation. Moreover, an action that is rational in the short term may create new risks in the long term or conflict with the goals of the organization as a whole.

Addressing these challenges requires technology that not only processes individual pieces of information but also relates changes in state, the results of actions, past experience, and higher-level goals to one another. TIGEREYE is researching and developing HSI, a proprietary AI model centered on state transition prediction and hierarchical decision-making, to realize AI that supports such real-world decision-making.

HSI is not intended simply to replace LLMs. As an AI model with a learning objective different from language generation, it aims to advance AI systems that support prediction and decision-making according to the situation, with future integration with existing AI technologies also in view.

■ The three core technologies of HSI

1. “State Learning,” which learns the relationships between states

HSI converts the multiple pieces of information that make up the current state into internal representations and learns the relationships among them. For example, for information such as work progress, priority, assignee, deadline, and the number of past responses, it learns how each affects the future state, using the prediction error for actions and next states. Its distinguishing feature is that it can form relationships useful for prediction inside the model, without people directly specifying the connections between pieces of information as the correct answer. Through this, it aims to acquire state representations that capture situations in which multiple factors influence one another and predict future changes.

2. “State Transition Prediction,” which predicts actions and future changes

From the current state, HSI predicts the next action to be taken and the next state that results from it. It is also designed to simulate future changes in state for multiple candidate actions and to evaluate reward, risk, constraints, future value, and more. For example, when changing the priority of work, it is intended to support choosing actions that take into account not only the immediate effect on work efficiency but also the impact on other processes and future risks. This builds a technology foundation for judging actions based on their future impact, not just their immediate results.

3. “Hierarchical Intelligence,” which evaluates the alignment between goals and actions

HSI handles goals, states, and actions at different levels of abstraction hierarchically. For example, it represents a company’s management goals, project goals, individual tasks, and actual actions as different levels, and evaluates the alignment between them. By analyzing changes in the degree of alignment, it is designed to detect deviations from goals, stalled work, signs of failure, and more, and to trigger warnings or changes in action when needed. This provides a mechanism for AI to choose actions based on their relationship to higher-level purposes, rather than focusing only on individual tasks.

■ “Experiential learning” that applies past experience to the next decision

In addition to state transition prediction and hierarchical decision-making, HSI incorporates a mechanism that applies past experience to future predictions.

Conventional information retrieval often emphasizes similarity, that is, how closely past information resembles the current situation. In real decision-making, however, even when a similar situation existed in the past, which experience is worth referring to differs depending on the action taken at that time, its result, when it happened, and the surrounding conditions.

HSI aims to make use of experience that is useful for the current state, taking into account not only the content of an experience but also its causal relevance, temporal proximity, the results of past actions, and more. Through this, it aims to realize AI that does not merely refer to past information but connects experience to predictions of future states and decisions on actions.

■ Basic verification of the proprietary AI model and future research and development

TIGEREYE is conducting basic verification of HSI’s core technologies using synthetic data and actual software development histories. To date, it has verified the model’s main functions, including learning the relationships among pieces of state information, predicting actions and next states, prediction that draws on past experience, and evaluating alignment across different levels. Through these efforts, it is examining the potential of a learning approach centered on state transitions and hierarchical decision-making.

HSI is currently in the research and development stage, and further verification is needed regarding its performance and versatility in real industrial environments. Going forward, TIGEREYE will continue research and development toward practical application, including verification using diverse real-world data, improving the model’s learning ability and prediction performance, and adapting it to different business environments.

■ Toward business development centered on a proprietary AI model

TIGEREYE aims to develop HSI not as an AI application limited to a specific use, but as a common AI model technology that can be applied across a variety of industries.

In the AI industry today, in addition to the development of foundation models centered on large language models, a wide range of research is under way, including robotics, autonomous AI agents, and World Models that predict changes in the state of the real world. In light of these technology trends, TIGEREYE will advance research and development of a proprietary AI model that integrates state transition prediction, experiential learning, and hierarchical decision-making.

Going forward, TIGEREYE plans to bring its research results into industry through integration with its proprietary AI platform “NERVA” and adoption support for companies from Forward Deployed Engineers (FDEs).

Expected application areas include the following.

• Autonomous AI agents: AI that supports goal-aligned choices of action based on changes in the situation and past experience
• Enterprise operations automation: More advanced business decisions that take into account progress, risk, and the goals of the organization as a whole
• Manufacturing: Technology that predicts changes in the state of equipment and processes to support production management and operational decisions
• Healthcare: Technology that supports healthcare professionals’ decisions using changes in state and historical information
• Security: Support for risk assessment and response decisions based on changes in the situation and signs of anomalies
• Robotics: Application to motion planning and action selection based on predictions of environmental states and the results of actions

TIGEREYE will also consider building new business models centered on AI models, such as licensing and API access for its proprietary AI model and expansion into industry-specific models. By advancing everything from AI research and development to implementation in companies as one, TIGEREYE aims to create social and business value from its proprietary technology.

■ Comment from Gaku Kamimura, CEO

“Generative AI has made it possible for people to draw on vast knowledge instantly. However, we believe the evolution of AI has only just begun.

What matters in the real world is not simply producing the right answer. It is understanding what is happening now, judging what should be done, and predicting how the future will change as a result of that action.

And every human and organizational activity has a purpose. Even if each individual action is correct, it will not necessarily lead to the best result unless it is aligned with the overall purpose.

Hierarchical State Intelligence, which we are developing, is a new approach that captures state, action, goals, and experience hierarchically so that AI can make decisions while predicting the future.

We want to be a company that does not merely use AI, but researches AI itself and creates its own intelligence technology.

TIGEREYE will take on the challenge of realizing a future beyond the era of AI that generates text: a future in which AI understands situations, makes decisions, and takes action.”

TIGEREYE Inc.
Gaku Kamimura, CEO

■ About the patent application

Item	Details
Technology name	Hierarchical State Intelligence (HSI)
Japanese name	階層型状態知能
Title of invention	状態情報を用いる学習モデルの情報処理方法、情報処理システム、及びプログラム
Applicant	TIGEREYE Inc.
Inventor	Gaku Kamimura
Application number	Japanese Patent Application No. 2026-226420
Filed with	Japan Patent Office
Technical field	Machine learning, state transition prediction, action prediction, hierarchical state learning, decision-making support
Application status	Patent application filed

This application concerns HSI’s core technology, including an information processing method that learns the relationships among pieces of state information and predicts actions and next states. TIGEREYE will continue research and development of its proprietary AI model and, through the protection and use of intellectual property, strengthen its technological competitiveness and drive expansion into industrial fields.

■ About TIGEREYE Inc.

AI as a force that moves society.
Intelligence into Action.

TIGEREYE Inc. is an AI technology company that drives everything from research and development of proprietary AI technology to product development and bringing AI into industry. It integrates AI technologies such as image recognition, language understanding, reasoning, judgment, and action, aiming to realize intelligence that functions in real work and in society.

Its business pillars include research and development of its proprietary AI model “Hierarchical State Intelligence (HSI),” the AI platform “NERVA,” and “Forward Deployed Engineers (FDEs),” who join the customer’s team and provide end-to-end support from discovering challenges to development, adoption, and embedding in operations.

Across industries including healthcare, manufacturing, offices, and hotels, TIGEREYE will evolve AI from a mere tool for operational efficiency into a presence that works alongside people and supports corporate growth and the creation of new value.

Create AI technology, deliver it as products, and solve real-world challenges.

TIGEREYE does not let the potential of AI end as a concept, but implements it as technology that truly functions in society.

Company name	TIGEREYE Inc.
Representative	Gaku Kamimura, CEO
Headquarters	Harumi Island Triton Square Office Tower W18F, 1-8-8 Harumi, Chuo-ku, Tokyo
Established	July 7, 2023
Business	AI model research and development, AI platform development, AI agent development, AI solution development, FDE business
Website	https://tigereye.ai/

Contact for this matter

Company name	TIGEREYE Inc.
Website	https://tigereye.ai/
Phone	03-6280-6762
Email	info@tigereye.ai