Self-developing AI service “Happiness Planet FIRA” goes live, promising to transform management discussions

Tokyo, August 26, 2025 – Happiness Planet, Ltd. (“Happiness Planet”) and Hitachi, Ltd. (TSE: 6501, “Hitachi”) have jointly developed a self-developing generative artificial intelligence (AI) capable of autonomously generating deeper insights and creative perspectives through discussions among hundreds of AI agents. Making use of the developed generative AI, Happiness Planet today began offering a self-developing AI service, “Happiness Planet FIRA”*1 (“FIRA”).

Conventional generative AI is limited in its ability to generate deep insights and creative perspectives that go beyond the provided training data. FIRA, applying proprietary technology from Happiness Planet and Hitachi, employs 600 types of AI agents specializing in various fields. Through autonomous discussion, they can generate deep insights and creative options geared to individual management issues without depending on user-specific data. As an intelligence amplifier*2 that expands human thinking, FIRA is applicable to a wide range of management situations, from devising management plans to handling investor relations (IR), coming up with new business ideas, formulating sales strategies, and improving the quality of management meetings. As an intelligent partner for senior executives and corporate planners, FIRA will contribute toward the sustainable growth of companies.

Demonstrating its potential in an objective evaluation on 10 management issues using AI, FIRA recorded significantly higher scores than conventional generative AI models.

*1

FIRA takes its name from Firenze (Florence), where the Renaissance began through interaction among various people, and arche, the first principle from which other principles arise.

*2

A concept, proposed by Doug Engelbart in 1962, of AI augmenting and expanding human intellect rather than replacing it.

Figure 1. Overall illustration of FIRA, in which 600 types of uniquely talented AI agents achieve self-development on the spot through intensive discussions with each other along the lines of the themes provided

Figure 2. Answers provided by major generative AI models (13 models in common use globally) and those provided by FIRA were objectively assessed on 10 management issues. The management assistance score is a benchmark where low marks are given to answers that anyone could come up with easily based on general thinking and high marks to those that deeply expand the user's thinking.

Background and issues

On a daily basis, people involved in corporate management or planning confront challenges that have no correct answer in areas such as risk management, investment decisions, and human resource development. These challenges require insight and creative perspectives tailored to the company's unique circumstances, going well beyond simply sorting through available information. However, conventional generative AI struggles to generate deep insights extending beyond the scope of pretraining data or provide creative perspectives tailored to the user's unique challenges. Its answers therefore tend to be limited to a general scope. Approaches to overcome this problem have been attempted by incorporating a company's past data into generative AI, but the reality is that simply referencing past data does not provide creative suggestions for the future.

FIRA features

Addressing such limitations, Happiness Planet and Hitachi developed AI technology that, without needing additional training data, generates deep insights and creative perspectives on the spot that are geared to the particular management issues that face the user, enabling creative answers tailored to unique circumstances (patent pending).

FIRA is designed to enhance the user's thought processes by providing a variety of perspectives as an intelligence amplifier that expands their knowledge and concepts. Serving as an intelligent partner for raising the quality of decision-making and maximizing creativity in corporate management and planning roles, FIRA supports diverse management issues. It can suggest new discussion points and perspectives in the drafting of management plans, for example, or offer multifaceted proposals for addressing management issues. It can also facilitate brainstorming at the concept stage of new business, research and development, or marketing discussions and help with developing human resources and leaders.

Specific features of FIRA include the following.

1. Uniquely talented AI agents capable of self-development

To realize an AI that could win out against a top Go player, it was initially necessary to train it on human game records. Later, however, the same feat was achieved with AI technology that develops on its own through trial and error without relying on prior data.*3 Applying this same approach to actual business challenges, though, with their vague rules and constraints, turned out to be more difficult.

Hitachi earlier developed self-developing AI agent technology applicable to specific problems, such as supply chain optimization and robots using a swing or horizontal bar.*4 Now, more universal application is possible with the newly developed self-developing agent technology through the use of large language models (LLMs). In the case of FIRA, these LLM-using self-developing AI agents are not ordinary agents; each has unique and exceptional capabilities. Drawing on their specialized perspectives and original thinking patterns, these agents autonomously offer opinions and proposals along the lines of themes or issues input by the user and discuss them among each other. A facilitator, in a command center role, exercises a certain amount of control over the direction of the discussions, the energy and emotional levels, the speakers, and the sequence and length of their statements. Under this control, the agents at times stimulate each other, at other times take opposing sides, or achieve cooperative harmony. In the process, they refine their queries and thinking, achieve self-development, and thereby generate deeper insights and creative options tailored to the unique circumstances of the company–all on the spot. In this way, unlike conventional generative AI, they are able to go beyond replies based on general theories and conventional wisdom to support a wide range of management challenges, such as drafting management plans, handling IR issues, or coming up with new business concepts.

*3

About AlphaZero: D. Silver et al., Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm

*4

About self-development by a robot learning to use a swing: https://www.youtube.com/watch?v=q8i6wHCefU4
About self-development by a robot working out on a horizontal bar: https://www.youtube.com/watch?v=uimyyGFwv2M
About self-development in the value chain: Hitachi, “Development of business-oriented AI technology based on competitive self-play learning,” news release, December 25, 2017.

2. Incorporates 600 kinds of uniquely talented, specialized AI agents equipped with the thinking of real-life experts

FIRA incorporates 600 types of AI agents, each with unique superpowers. They were designed by Kazuo Yano (Representative Director and CEO of Happiness Planet and Fellow of Hitachi), whose research on combining AI and human science has won a number of international awards.*5 Among the agents are a “Super CFO” with specialized knowledge in financial affairs, “Dr. Innovator,” whose specialty is coming up with game-changing ideas, and many other independent AI agents possessing specialized knowledge and values in fields such as business administration, economics, psychology, IR, finance, human resources, data science, concept-making, wellbeing, history, science, thinking, and philosophy. In addition to these are Bunshin (avatars)*6 equipped with the thinking of real-life experts (mental trainer Asami Kuru, Bears Executive Vice President Yuki Takahashi, athlete Dai Tamesue, writer Shu Yamaguchi, and Kazuo Yano). These diverse, uniquely talented agents, carrying on intensive discussions facilitated by AI, are able to generate new ideas and wide-ranging thinking that conventional AI cannot produce, expanding the potential of business management.

*5

2020 IEEE Frederik Phillips Award, 2007 Mind, Brain, and Education Erice Prize, etc.

*6

Happiness Planet, “Happiness Planet develops new technology merging the thinking of top runners in various fields with generative AI: Launching new application Happiness Planet/Regrowth, an AI counselor for the user's everyday concerns” [in Japanese], news release, March 21, 2024.

3. Objective verification of FIRA management assistance capabilities

The performance of FIRA was compared to that of other major generative AI models (13 models in common use globally) based on a “management assistance score” benchmark. This benchmark assigns the highest scores not to answers that a top executive could come up with easily based on general thinking but to those that deeply expand the user's thinking. The specific evaluation approach, called the LLM-as-a-Judge method,*7 consists of an objective, blind assessment of the answers given by each AI on 10 management themes (risk, finance, investment, human resources, technology, marketing, transformation, etc.). The answers are assigned points by other generative AI models on a scale from 1 to 8 for each of seven evaluation criteria (surprise, depth, metacognition, etc.). In this comparison, FIRA significantly outperformed conventional major generative AI models. Whereas the major generative AI models received scores that were from “somewhat low” to “somewhat high,” FIRA scored close to “very high,” recording a T-score 27 points higher above the average conventional model score.*8 This result suggests that FIRA has promising potential to contribute significantly to both raising the quality of management decision-making and new value creation.

*7

A method whereby qualitative evaluations of answers are given by the LLM automatically, without relying on human judgment.

*8

In this evaluation, the AI answers on 10 themes were assessed on a scale of 1 to 8 by two models, ChatGPT-4o and o3, with the mean of the scores by the two models used as the overall assessment.

Looking ahead

Applying the management co-creation that brings people and AI together, made possible by the self-developing generative AI they have developed, Happiness Planet and Hitachi aim to support the transformation of companies and organizations. The main focus of this support, by means of AI that redefines queries and expands knowledge, is on assisting people with the process of constantly reframing the situation so that they can make better choices in context. FIRA is scheduled to be demonstrated at the AI expo on August 27–28, 2025, at the Tokyo International Forum.

Figure 3. A network diagram showing the relationships among 600 kinds of uniquely talented AI agents. The nodes represent agents, and the colors show the fields of their expertise. Based on a quantitative assessment of the relationships among agents, those with relatedness above a certain level are connected by lines and are arranged in closer proximity, resulting in this network shape. The AI agents specializing in management-related fields are spread throughout the entire diagram.

About Happiness Planet, Ltd.
Happiness Planet provides AI and services with the express aim of connecting happiness to human and organizational productivity. It was founded in July 2020 as a Hitachi Group company with the mission of utilizing human data and AI to create a new industry. Based on a scientific understanding of people and organizations derived from large amounts of data, the company provides AI services supporting the building of happy and creative enterprises. For more details, see the company website (https://happiness-planet.org/ [in Japanese]).

About Hitachi, Ltd.
Through its Social Innovation Business (SIB) that brings together IT, OT(Operational Technology) and products, Hitachi contributes to a harmonized society where the environment, wellbeing, and economic growth are in balance. Hitachi operates globally in four sectors – Digital Systems & Services, Energy, Mobility, and Connective Industries – and the Strategic SIB Business Unit for new growth businesses. With Lumada at its core, Hitachi generates value from integrating data, technology and domain knowledge to solve customer and social challenges. Revenues for FY2024 (ended March 31, 2025) totaled 9,783.3 billion yen, with 618 consolidated subsidiaries and approximately 280,000 employees worldwide. Visit us at www.hitachi.com.

Inquiries
Happiness Planet, Ltd.
Kazunori Yanagi, SHUN Tanoue
Inquiry Form:

Hitachi, Ltd.
Research & Development Group
Inquiry Form: