Home
A challenge of current generative AI

Fujitsu's autonomous AI agent

Values of our AI agent
- Quick installation without expertise
- It will become possible to quickly make a decision at better quality. Users can use immediately right after the installation, and it will be smarter by using it on regular basis. No expertise required.
- Suggesting solutions based on context understandings
- Based on the understanding of the context from the past transactions and meeting materials, AI agent autonomously analyze ongoing discussions and show the solutions.
- Suggesting a variety of approaches to subjects
- There can be several solutions to a subject that AI agent found. Autonomous learning technology in AI agent can suggest users a variety of approaches including ones out of context. The feature help users enhance creativity.
|
|
|
|
|
|
Overview of technology
Target businesses and users
Initially it targets board members who want to make a decision in the management meetings. Subsequently it will be offered as a tool for communication support at enterprize in wide range.
Subjects in targeted businesses and operations
Enterprises intend to analyze in-house data by utilizing AI technology in decision making process. However, it is not easy because advanced expertise is required to find out the best combination of AI technology and data. In such a circumstance, the technology of AI agent, leveraged by the power of Large Language Models (LLMs), is in demand that can autonomously combine AI and data in building up solutions.
Technical subjects in the AI agent technology
- A task can be generated for a topic in discussion, but it is not often aligned with the context of meeting. Therefore, the contribution is limited in practice.
- A straightforward task can be generated aligned with discussion, while it is hard to generate tasks beyond the mind of attendees. Therefore, attendees may not notice substantial sunjects or new ideas behind ongoing discussions.
Values in detail that Fujitsu's autonomous AI agent offers
- Before the introduction of AI agent: AI cannot work without prompts given by humans. In meeting, it is quite possible that attendees themselves have not recognized the subjects. In such cases, attendees cannot prompt AI that results in the limited usage of AI capabilities.
- After the introduction of AI agent: By just inviting AI agent to the meetings, it autonomously understands the context of discussions, generate tasks, and execute them to solve abstract subjects. Attendees can realize new insights by AI agent that is autonomously give them information or suggest them provisions, it becomes possible to smoothly conduct meeting and obtain productive conclusion.
Superiority of Fujitsu's technology
- Context retention technology that facilitates AI agent to propose subjects and solutions based on the past minutes and the meeting materials.
- Self-learning technology that enumerates a variety of interpretations on the same subject.
- Uniquely developed AIs including "Takane", an LLM that has the highest Japanese language capability, and Fujitsu AutoML, an AutoML capable of generating highly accurate machine learning pipelines. Not only our AI technologies, other AIs developed by other vendors can be combined.
- It is easy to build AI agent into Microsoft products by utilizing Microsoft coalition (Semantic Kernel, Copilot etc).
Usage scenes
- End user
- In a wide range of the meetings such as board meeting, or profit-and-loss meeting, or project meeting, or 1-on-1 etc, it can be utilized as a tool for communication support. Attendees can obtain realizations and inspirations. Productivity of the meeting is improved.
Use cases
In practice at Fujitsu now.
Trial
The below video shows our AI agent demonstrated in Strategy Briefing on 12th in December in 2024.
It is possible to conduct a PoC using our technology.
Related information
- Press release
- Other information
Documents
In preparation