Kunihiko Harada
Research & Development Group
Hitachi, Ltd.
Introduction
In today’s fast-paced business environment, leveraging data-driven decision-making is no longer optional – it is a necessity. In this blog, I’d like to introduce a graph-based framework developed to enhance organizational intelligence and support continuous business improvement by my colleagues and me in Hitachi R&D. By combining cutting-edge generative AI and graph neural networks (GNNs), this framework bridges the gap between management strategies and frontline execution, which empowers businesses to adapt to dynamic environments.
Role of knowledge graphs
Knowledge graphs are versatile tools that organize and interlink diverse data points into a meaningful network of relationships. They enable businesses to uncover patterns, optimize processes, and drive strategic decision-making. Our team developed a framework that organizes organizational challenges, operational workflows, and technological insights into four systematic tasks:
- Stakeholder analysis: Identifying key stakeholders and mapping their relationships.
- Challenge identification: Visualizing management, operational, and technological challenges.
- Factor identification: Analyzing processes and data for actionable insights.
- Methodology design: Designing data-driven strategies for implementation.
By adopting this structured approach, organizations can address challenges holistically and ensure alignment between high-level goals and day-to-day operations.
Applications
Our framework integrates domain expertise and AI techniques, which makes it more powerful and accessible. Its key applications include:
- Automatic generation of challenge graphs
Using large language models (LLMs), we extract and organize organizational challenges into comprehensive “challenge graphs.” These serve as the foundation for aligning business objectives with solutions, simplifying the often-complex process of pinpointing core challenges. - Graph review assistance
Understanding that initial graphs may lack key operational details, this feature suggests additional nodes and connections – using insights from historical data – to refine and enrich the graph’s accuracy. - Solution design and node scoring
Finding the most suitable solutions to business challenges can be time-consuming. By employing GNNs, we highlight high-potential solutions while node scoring helps prioritize the most impactful challenges with minimal user input.
By unifying unstructured and structured data in an app-driven human-in-the-loop approach, we catalyze a continuous knowledge-accumulation cycle, which fosters organizational intelligence.
Practical impact
Our trial upon one of the applications revealed significant benefits. By automating repetitive tasks such as market analysis, the application saved approximately 83% of the workload in time, which enabled teams to focus on strategic activities. Furthermore, the framework facilitated deeper understanding of industry trends.
Conclusion
Knowledge graphs are shaping the future of organizational intelligence. By integrating generative AI and GNNs, our framework transforms fragmented data into actionable insights. We invite you to explore these technologies and join us in the journey toward smarter data-driven decision-making.
Acknowledgements
I would like to thank Mr. Yajima and Dr. Matsumoto, who co-authored the research work.
Reference
[1] Y. Yajima, S. Matsumoto, and K. Harada, “Knowledge graph data analysis for organizational intelligence to support continuous business improvement,” Connected Data London 2024, 2024.
