According to @DeepLearningAI, SAP's Christoph Meyer and Lars Heling detailed at AI Dev 25 x NYC how knowledge graphs improve AI agent discovery and execution by providing semantic and process context to safely invoke the right tools and enterprise APIs, with LLMs supplying fluency and knowledge graphs supplying effectiveness; they covered semantic retrieval, process-aware API connectivity, alignment with the Model Context Protocol (MCP), and ran a demo applying these methods, with the full talk available at piped.video/watch?v=XrYwFUdu2lk; source: @DeepLearningAI on X, Dec 18, 2025. The post does not mention cryptocurrencies, tokens, or market impacts, indicating the session focused on enterprise AI agent reliability rather than crypto-specific use cases; source: @DeepLearningAI on X, Dec 18, 2025.