[Joyo Bank] Launch of Business-Specialized AI Agent "JOYO AI AGENT" for Internal Use
Joyo Bank has begun internal use of its business-specialized AI agent "JOYO AI AGENT," a locally-run LLM operating in a closed environment completely isolated from the internet.
📋 Article Processing Timeline
- 📰 Published: March 30, 2026 at 00:26
Joyo Bank (President: Tetsuya Akino) has built a business-specialized AI agent (local LLM) called "JOYO AI AGENT" that operates in a closed-network environment completely isolated from the internet, and announced the start of its internal use on March 26, 2026.
Since July 2025, the bank has been conducting proof-of-concept trials in collaboration with Athena Technologies, Inc. (Representative Director: Takeshi Abe), a startup originating from the Matsuo Laboratory that develops generative AI solutions. The trials confirmed the system's effectiveness in enhancing and streamlining operations while protecting the bank's confidential information and customers' personal data, leading to the decision to adopt the system.
The bank will continue to leverage the latest technologies to improve management efficiency and establish a stable business foundation, while contributing to the sustainable growth of the local community by sharing the knowledge gained through internal use.
---
1. Background and Features of "JOYO AI AGENT"
In recent years, there has been growing demand for AI-driven efficiency improvements in banking operations, while ensuring the safety of data containing confidential and personal information has become a critical challenge.
"JOYO AI AGENT" builds the generative AI system environment on a closed cloud that is physically separated from the external internet, reducing data leakage risks and achieving a high level of security. It is also designed with scalability to accommodate additional use cases, as multiple agents specialized for different business scenarios are implemented.
The system also features a "prompt-less" (no instruction text required) user interface where processing is executed simply by uploading a file and clicking a button. This allows bank employees without specialized AI knowledge to use the system smoothly.
2. Implemented Features (Use Cases)
In this rollout, alongside the AI agent infrastructure, the following two functional areas were implemented based on use cases validated during the proof-of-concept trials. Going forward, the bank will continue to expand available use cases across various departments, including branch offices.
| Feature | Status | Overview |
|---|---|---|
| Translation | Released March 2026 | Performs translation (English↔Japanese) of highly specialized documents such as financial texts, capturing technical terminology and context. |
Since July 2025, the bank has been conducting proof-of-concept trials in collaboration with Athena Technologies, Inc. (Representative Director: Takeshi Abe), a startup originating from the Matsuo Laboratory that develops generative AI solutions. The trials confirmed the system's effectiveness in enhancing and streamlining operations while protecting the bank's confidential information and customers' personal data, leading to the decision to adopt the system.
The bank will continue to leverage the latest technologies to improve management efficiency and establish a stable business foundation, while contributing to the sustainable growth of the local community by sharing the knowledge gained through internal use.
---
1. Background and Features of "JOYO AI AGENT"
In recent years, there has been growing demand for AI-driven efficiency improvements in banking operations, while ensuring the safety of data containing confidential and personal information has become a critical challenge.
"JOYO AI AGENT" builds the generative AI system environment on a closed cloud that is physically separated from the external internet, reducing data leakage risks and achieving a high level of security. It is also designed with scalability to accommodate additional use cases, as multiple agents specialized for different business scenarios are implemented.
The system also features a "prompt-less" (no instruction text required) user interface where processing is executed simply by uploading a file and clicking a button. This allows bank employees without specialized AI knowledge to use the system smoothly.
2. Implemented Features (Use Cases)
In this rollout, alongside the AI agent infrastructure, the following two functional areas were implemented based on use cases validated during the proof-of-concept trials. Going forward, the bank will continue to expand available use cases across various departments, including branch offices.
| Feature | Status | Overview |
|---|---|---|
| Translation | Released March 2026 | Performs translation (English↔Japanese) of highly specialized documents such as financial texts, capturing technical terminology and context. |