Customer Case Study: Fujitsu Kozuchi AI Agent Powered by Semantic Kernel

Japanese multinational Fujitsu, a pioneer of information and communications technology, has been transforming industries since 1935. With a workforce of 124,000 dedicated professionals across 50 countries, Fujitsu’s mission is to “make the world more sustainable by building trust in society through innovation.”

After introducing their groundbreaking Composite AI earlier this year, Fujitsu is back with another innovative business solution. The Fujitsu Kozuchi AI Agent is an autonomous system that functions as a full team member, making proactive suggestions and analyses. 

Launched in late October for use in business profitability discussions and negotiations, a broader line of agents will eventually expand to cover production management and legal affairs, among other functions. 

From Interactive to Proactive

Early generative AI was based on a call-and-response model that required human users to input instructions and only generated responses based on previously learned data. These models were useful tools that increased productivity but lacked the ability to discover issues on their own or propose novel solutions.

The Fujitsu Kozuchi AI Agent uses proprietary processing logic to break down abstract issues from conversations and generate discrete tasks. It proactively enters and participates in a discussion, creates a plan to complete tasks, and selects multiple AI models to implement. 

It acts just as a regular team member would, making proactive suggestions and generating new, creative ideas. 

Practical example

Let’s say two team members are discussing the current sales figures for an important region, and there’s a disagreement over performance. The meeting transcription reads like this:

Team member 1: Recently, I’ve been hearing that deals in Asia are not going well.

Team member 2: No, I think Asia is fine. I’ve heard we lost a few small deals, but I don’t think it’s affecting the big picture.

Team member 1: I know Dan, the head of Asia, has said as much, but apparently, the numbers show otherwise. 

With the AI agent present, additional data could be suggested to clear up the confusion. The system would select several AI models and direct them to perform sales analysis in the Asia region. Depending on the result, it would present a clear answer, easily digestible visuals, and reasons for the discrepancy.

This is done completely autonomously, without either team member inputting instructions or interacting with the AI. The meeting can continue with better insight, and decisions can be made more effectively.

Fujitsu Kozuchi AI Agent

Key features of Fujitsu Kozuchi AI Agent

Available as part of the Fujitsu Uvance business model, the Fujitsu Kozuchi AI Agent leverages the power of multiple AI models, including:

• Takane: Large-scale language model for enterprises with the world’s best Japanese language capability and significant customization options.
• Fujitsu AutoML: Generates advanced machine learning models in a short amount of time.
• Conversational Optimization: Automatically formulate, solve, and visualize optimization problems through chat-style input

The AI Agent provides instructions to each model and provides a solution based on the feedback. The Semantic Kernel SDK powers the agent and allows it to choose the optimal AI service for each task.

As more agents are released for other business functions like production management and legal affairs, the Kozuchi ecosystem will work together to provide more accurate data, decisions, and suggestions.

Summary

Semantic Kernel’s robust enterprise-level development kit has allowed Fujitsu to continue building AI business applications that previously required a significant amount of time, money, and expertise. The Fujitsu Kozuchi AI Agent makes proactive suggestions to inspire creativity. Fujitsu will present on this during Microsoft Ignite in this session: Productive AI with Semantic Kernel. Fujitsu respects human-centric AI. With the Fujitsu Kozuchi AI Agent, we are mindful of ensuring that it collaborates with humans without hindering their decision-making capabilities or autonomy.

References

Learn more about Fujitsu, the Kozuchi AI ecosystem, and Uvance. For more information about the Fujitsu Kozuchi AI Agent, check out the full release.

Learn more about Semantic Kernel.