Causal AI technology for explainable and actionable decision-making

Fujitsu Causal AI is a decision intelligence technology that helps organizations identify the factors that drive business outcomes and recommend actions optimized to improve those outcomes. Unlike conventional analytics, which focuses primarily on correlations, Fujitsu Causal AI applies causal analysis to enable explainable, evidence-based, and actionable decision-making.

Methodology

The technology applies causal discovery techniques to observational data to estimate potential cause-and-effect relationships and visualize them as causal graphs. It then uses the resulting causal models to evaluate and optimize actions that can move a target KPI from its current state toward a desired state while accounting for operational constraints. Domain knowledge and prior causal knowledge, including previously identified causal relationships and causal graphs, can also be incorporated to improve the reliability of the analysis, particularly when the available data are limited. The technology supports natural-language interaction for business users and is designed to provide multiple access modes, including advanced analytical capabilities for specialists, SDK-based integration, and agent-to-agent interfaces. It can identify causal factors and recommend mathematically optimized actions while providing an explainable rationale for those recommendations.

Performance and Evidence

Fujitsu has developed LayeredLiNGAM, a high-speed causal discovery method that can perform causal discovery up to 1,000 times faster than DirectLiNGAM under the benchmark conditions described in Fujitsu's published materials. The technology has been applied or evaluated across multiple domains, including manufacturing yield improvement, healthcare, personalized recommendations, ESG analysis, and financial performance management.

Applications

Potential applications include:

• Optimization of pricing, promotions, inventory, staffing, allocation, and process settings
• Manufacturing quality and yield improvement
• Healthcare and lifestyle intervention planning
• Personalized retail offers, coupons, and recommendations
• ESG and sustainability analysis
• Identification of financial and non-financial drivers of business performance

Proposed Partnership   We are interested in collaborating with software companies, solution providers, and enterprises seeking to evaluate, integrate, or apply Fujitsu Causal AI technology. Potential collaboration models include technical evaluation, proof-of-concept projects, solution integration, co-development, and, where appropriate, technology licensing.

Proposed Next Steps

• Initial technical and business discussion
• Identification of a suitable use case and target KPI
• Assessment of available data and operational constraints
• Trial or proof of concept
• Consideration of production deployment, solution integration, or licensing