A Q&A system using AI that can quickly acquire advanced expertise

Furuno Electric Co., Ltd. develops a wide range of businesses, including the manufacture and sale of electronic equipment for ships, as well as healthcare and information communication. Furuno Electric engineers are active in locations around the world, promoting business activities to contribute to customers and society. However, they were experiencing issues with passing on advanced specialized skills from veteran engineers to young engineers.

We used this opportunity to develop AI Support, a highly specialized Q&A system that provides on the spot support for technology used by engineers responding to calls in the field.

This time, in order to make further use of this system, we launched a renewal project aimed at improving the scalability and response accuracy. For this project, Fenrir was in charge of the development phase.

With the goal of “creating a culture where engineers from all locations can use AI Support to share the knowledge of engineers from all over the world”, we created an environment where Furuno Electric employees can quickly obtain the necessary information for troubleshooting and problem solving.

AI Support received the Data Management Award 2025 from the Japan Data Management Consortium (JDMC), which recognizes companies and organizations that are practicing exemplary activities in data management. AI Support was evaluated highly for providing solutions to the problem of the shortage of mid-level human resources faced by many companies. It has also gained attention in the industry as an AI technology that solves the problem of inheritance of technology between generations.

Consultation points

Challenges we addressed

Secure data management environment

In order to realize technology transmission through AI Support, we utilized a huge amount of internal documents and knowledge, which are data sources. Designed with security in mind to ensure the safety of data management in handling data, which can be said to be corporate assets. We adopted Okta, a certification service used by Furuno Electric, so that only a limited number of users can access and use the system.

Both customization and ease of operation

Furuno Electric was looking for customization of the AI Support system.

In addition to making it easy to customize, we implemented infrastructure design that is easy to manage and operate afterwards. The configuration is scalable, flexible, and designed to support future data growth.

AI support is operated in a cloud environment based on AWS. We adopted a design that improves availability by AWS multi-AZ configuration, minimizing the impact on site use even in the event of a failure, and achieved stable operation. In addition, we adopted a design that utilizes fully managed services such as App Runner and DynamoDB to minimize the burden of system management and operation and maintenance as much as possible.

In addition, by building CI/CD, we have automated manual labor such as deploying applications that occur in the development process. In the future, developers will be able to improve the efficiency of their work when developing additional products.

Utilizing OpenSSA for system optimization

In response to the questions of the engineers of Furuno Electric, Fenrir chose OpenSSA to output technical information.

OpenSSA is an open source framework developed by AITOMATIC to develop small-scale specialized AI agents (SSAs) for specific industries.

Fenrir collaborated with AITOMATIC to develop and provide AI solutions for businesses using OpenSSA.

A systematic HTP plan breaks down complex tasks on an easy-to-handle scale and leads to a solution by placing them in the OODA loop, a framework for rapid decision-making.

In introducing OpenSSA, we promoted the business understanding of the engineers of Furuno Electric who actually use it with the goal of effective use of functions. On top of that, we repeatedly discussed how to use OpenSSA, and adjusted it assuming that it will be used in the field, such as minimizing the number of inferences in order to reduce the response time without using the default function of OpenSSA as it is.

In addition, OpenSSA introduced a multi-agent architecture that uses multiple agents. By selecting agents specializing in business and specific fields, we have built a system that can respond appropriately to that field.

On the other hand, there was also a problem when it was assumed that OpenSSA should be used with Azure OpenAI. OpenSSA does not support Azure OpenAI by default, and the accuracy is high, but the answer speed is slow. Therefore, it is suitable for AI support by analyzing the source code of OpenSSA and implementing the corresponding code so that OpenSSA can also be used in Azure OpenAI.

Internal Q&A system using internal documents

In AI Support, we utilized RAG* technology.

Since AI searches for pre-specified internal documents and responds, it is now possible to provide accurate and consistent information. In addition, we ensure the reliability of the answer by displaying the materials, pages, and URLs referenced at the time of the answer.

In addition, you can choose an AI model from Azure OpenAI, AWS Bedrock, and OpenSSA according to the content of the question, so you can adjust how far you can freely think about AI.

In addition, the UI is multilingual. At the time of the inquiry, we chose to focus on the English, Japanese, and Chinese languages, and ensured that answers would be provided in the same language as the inquiry.

Add AI Support as a chatbot to Teams.

We provided an environment where you can work on your daily work as if you were having a technical-related employee next to you.

*RAG: Retrieval-Augmented Generation

This is a process to optimize the output of large-scale language models (LLMs). The combination of the powerful capabilities of LLM with the internal knowledge base of a particular field or organization enables more relevant, accurate, and useful output. It is a cost-effective approach because there is no need to retrain LLM, and it can respond to various situations.

Future outlook

AI Support, which was able to respond specifically to the field by designing the strengths of AI models, has become a reassuring presence that supports the work of engineers.

In the future, with the aim of realizing a long-term vision of “creating a culture where engineers from all bases can share the knowledge of engineers from all over the world using AI Support” will be realized, and we will absorb the feeling of system use and continuously update it.