AWS re:Invent 2025 - Knowledge Graphs for AI and Intelligent Systems (DAT209)

Summary of "Knowledge Graphs for AI and Intelligent Systems" (AWS re:Invent 2025)

Challenges with Enterprise Data for AI

• Only 5% of AI projects make it to production, with the main reason being that enterprise data is not "AI-ready"
• Key challenges include:
  • Data silos, security constraints, and regulatory issues
  • Hallucinations and lack of transparency in AI models
  • Lack of structured, connected, and context-rich data for language models

Importance of Context Engineering

• Context engineering is an evolution of prompt engineering, providing the right information at the right time to AI models
• Key sources of context include user interactions, model memory, structured APIs, and knowledge graphs
• Providing the "minimal viable context" is crucial to avoid overwhelming AI models with too much information

Knowledge Graphs for AI Readiness

• Knowledge graphs are a way to organize and access connected data, using a property graph model of nodes and relationships
• Knowledge graphs solve the "query diversity" problem by providing a consolidated, connected view of data
• They also enable richer context to be provided to language models, improving accuracy and explainability

Technical Details and Architecture

• Knowledge graphs can be integrated as a "context layer" between enterprise data platforms and language models
• Graph databases enable efficient querying, pattern matching, and path-finding algorithms
• Studies show knowledge graphs can provide 3x more accurate responses from language models compared to SQL/NoSQL databases

Business Applications and Impact

• Knowledge graphs enable more trustworthy and explainable AI systems, especially for complex enterprise use cases like supply chain optimization
• The connected, context-rich data in knowledge graphs allows language models to reason about multi-step processes and make more informed decisions
• Visualizing and exploring knowledge graphs provides additional business value beyond just powering AI applications

Key Takeaways

• Knowledge graphs are a critical enabler for making enterprise data "AI-ready" by providing structured context
• Integrating knowledge graphs as a context layer can significantly improve the accuracy, explainability, and business impact of AI systems
• Knowledge graphs go beyond just powering graph-based retrieval, enabling a wide range of advanced AI and analytics capabilities