Graat Supports Planning and Requirements Definition Reform Utilizing AI Agents in au PAY's In-house Agile Development
Graat, a subsidiary of Growth ExPartners, has collaborated with KDDI to support the reform of planning and requirements definition processes in KDDI's in-house agile development for its QR code payment service "au PAY," by leveraging AI agents. This initiative integrates AI agents into the planning and requirements definition workflow, incorporating the judgment criteria and procedural know-how of experienced personnel in a reusable format. This enhances the quality of communication between planning and development, making it easier to dedicate time to discussions on essential requirement values and appropriate implementation methods. This case is a representative example of Graat's AI Agent Design Service "EBAAD (Enterprise Business AI Agent Design)."
📋 Article Processing Timeline
- 📰 Published: April 27, 2026 at 17:00
- 🔍 Collected: April 27, 2026 at 08:31
- 🤖 AI Analyzed: April 27, 2026 at 09:46 (1h 14m after Collected)
Growth ExPartners Inc. (Headquarters: Shinjuku-ku, Tokyo; Representative Director and President: Shinichi Watanabe)'s subsidiary, Growth Architecture & Teams Inc. (Headquarters: Shinjuku-ku, Tokyo; Representative Director and President: Yusuke Suzuki; hereinafter "Graat"), in collaboration with KDDI Corporation (Headquarters: Minato-ku, Tokyo; Representative Director and President: Koji Matsuda; hereinafter "KDDI"), is supporting the reform of planning and requirements definition utilizing AI agents in KDDI's in-house agile development for its QR code payment service "au PAY."
In this initiative, AI agents are incorporated into the series of planning and requirements definition processes, and the judgment criteria and procedural know-how of experienced personnel are reflected in a reusable format. This has improved the quality of communication between planning and development, making it easier to dedicate time to discussions on essential requirement values and appropriate implementation methods.
This case is a representative practical example of Graat's AI Agent Design Service "EBAAD (Enterprise Business AI Agent Design)*."
About "EBAAD": https://www.graat.co.jp/genai
EBAAD is a consulting service that integrates the judgment and response know-how of skilled personnel accumulated within companies into business processes through AI utilization. It transforms individualized knowledge into a mechanism that continuously generates value, designing and operating it in a state that can be utilized by the organization. Instead of merely introducing generative AI and AI agents as tools for operational efficiency, it re-examines the flow of operations and division of roles, and by continuously enabling AI to function, it supports the improvement of corporate organizational capabilities and the strengthening of sustainable competitiveness.
Background: Individualized expertise in planning and requirements definition led to quality variations and increased coordination burden.
KDDI has been promoting initiatives to utilize AI in its development processes as it advances in-house agile development for au PAY. However, in the area of planning and requirements definition, the judgment criteria and procedural know-how held by skilled personnel and developers tended to be individualized, leading to a challenge of variations in the quality of deliverables depending on the person in charge.
In planning and requirements definition, it is necessary to organize "what should be achieved" and "how it should be appropriately achieved" based on business requirements, system impact, and stakeholder perspectives. However, many of these judgments are supported by the tacit knowledge of experienced personnel, making it difficult to reproduce across the entire organization through documentation and training alone.
To address this challenge, Graat is supporting the transformation of individualized knowledge into a form that can be utilized by the organization, not by merely introducing AI as a tool or for operational efficiency, but by re-examining the approach to planning and requirements definition itself and continuously integrating AI agents into the workflow.
Overview of the Initiative: Collaborative design and implementation of a planning and requirements definition support mechanism incorporating AI agents.
Graat collaborated with KDDI's Personal Business Division, System Development Department, Agile Development Section, to design, implement, and provide ongoing support for the operation and improvement of a "Backlog Creation Support System" – a mechanism to assist planning and requirements definition in in-house agile development – targeting multiple teams involved in au PAY.
The main focus of this initiative is to re-evaluate the division of roles between humans and AI across the entire planning and requirements definition process, and to design business processes where AI functions continuously. Graat provides consistent support from concept organization, business visualization, process design based on AI, system implementation, to operational improvement on-site.
Features of the Initiative: Combining thought support and task delegation to enhance the quality of requirements definition.
Graat emphasizes designing AI utilization not as simple automation, but as a combination of "thought support" and "task delegation." In planning and requirements definition, it is crucial that AI does not decide requirements on behalf of humans, but rather supports humans in considering value and implementation methods, and then delegates tasks.
For example, the Backlog Creation Support System provides the following support:
1. Requirements Validation
Supports checking the relevance of background, objectives, and means, and verifying the lack of necessary information and logical consistency.
2. Supplementing Necessary Knowledge
Refers to related existing systems and peripheral information, supplements necessary prerequisite knowledge for consideration, and promotes planning and requirements definition while compensating for differences in担当者 experience.
3. Stakeholder Perspective Review
Conducts reviews simulating multiple perspectives such as end-users, engineers, and business personnel, promoting consideration of issues that might not be easily noticed by the person in charge alone.
4. Backlog Generation and Registration
Generates a draft backlog based on organized requirements and peripheral information, and supports registration into ticket management tools.
In this initiative, AI agents are incorporated into the series of planning and requirements definition processes, and the judgment criteria and procedural know-how of experienced personnel are reflected in a reusable format. This has improved the quality of communication between planning and development, making it easier to dedicate time to discussions on essential requirement values and appropriate implementation methods.
This case is a representative practical example of Graat's AI Agent Design Service "EBAAD (Enterprise Business AI Agent Design)*."
About "EBAAD": https://www.graat.co.jp/genai
EBAAD is a consulting service that integrates the judgment and response know-how of skilled personnel accumulated within companies into business processes through AI utilization. It transforms individualized knowledge into a mechanism that continuously generates value, designing and operating it in a state that can be utilized by the organization. Instead of merely introducing generative AI and AI agents as tools for operational efficiency, it re-examines the flow of operations and division of roles, and by continuously enabling AI to function, it supports the improvement of corporate organizational capabilities and the strengthening of sustainable competitiveness.
Background: Individualized expertise in planning and requirements definition led to quality variations and increased coordination burden.
KDDI has been promoting initiatives to utilize AI in its development processes as it advances in-house agile development for au PAY. However, in the area of planning and requirements definition, the judgment criteria and procedural know-how held by skilled personnel and developers tended to be individualized, leading to a challenge of variations in the quality of deliverables depending on the person in charge.
In planning and requirements definition, it is necessary to organize "what should be achieved" and "how it should be appropriately achieved" based on business requirements, system impact, and stakeholder perspectives. However, many of these judgments are supported by the tacit knowledge of experienced personnel, making it difficult to reproduce across the entire organization through documentation and training alone.
To address this challenge, Graat is supporting the transformation of individualized knowledge into a form that can be utilized by the organization, not by merely introducing AI as a tool or for operational efficiency, but by re-examining the approach to planning and requirements definition itself and continuously integrating AI agents into the workflow.
Overview of the Initiative: Collaborative design and implementation of a planning and requirements definition support mechanism incorporating AI agents.
Graat collaborated with KDDI's Personal Business Division, System Development Department, Agile Development Section, to design, implement, and provide ongoing support for the operation and improvement of a "Backlog Creation Support System" – a mechanism to assist planning and requirements definition in in-house agile development – targeting multiple teams involved in au PAY.
The main focus of this initiative is to re-evaluate the division of roles between humans and AI across the entire planning and requirements definition process, and to design business processes where AI functions continuously. Graat provides consistent support from concept organization, business visualization, process design based on AI, system implementation, to operational improvement on-site.
Features of the Initiative: Combining thought support and task delegation to enhance the quality of requirements definition.
Graat emphasizes designing AI utilization not as simple automation, but as a combination of "thought support" and "task delegation." In planning and requirements definition, it is crucial that AI does not decide requirements on behalf of humans, but rather supports humans in considering value and implementation methods, and then delegates tasks.
For example, the Backlog Creation Support System provides the following support:
1. Requirements Validation
Supports checking the relevance of background, objectives, and means, and verifying the lack of necessary information and logical consistency.
2. Supplementing Necessary Knowledge
Refers to related existing systems and peripheral information, supplements necessary prerequisite knowledge for consideration, and promotes planning and requirements definition while compensating for differences in担当者 experience.
3. Stakeholder Perspective Review
Conducts reviews simulating multiple perspectives such as end-users, engineers, and business personnel, promoting consideration of issues that might not be easily noticed by the person in charge alone.
4. Backlog Generation and Registration
Generates a draft backlog based on organized requirements and peripheral information, and supports registration into ticket management tools.