Steal This Deck

2026-05-09 · Source: Intentional Arrangement · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Intermediate, long

Summary

This article recaps a Knowledge Graph Conference 2026 presentation titled "Stop Betting, Start Building," later renamed "Steal This Deck," which argues that reliable agentic AI requires robust knowledge infrastructure, not just data. The author contends that despite market enthusiasm, current AI applications show minimal productivity impact, with 89% of firms reporting no gains and experienced developers being slower with AI tools. The core argument is that AI functions as a knowledge tool, not merely a data tool, evidenced by large language models (LLMs) being trained on knowledge-rich, linked-data sources like Wikipedia and Google Patents. The presentation outlines a six-layer knowledge infrastructure progression from controlled vocabularies to knowledge graphs, emphasizing that skipping layers leads to brittle, opaque, and inefficient AI systems prone to hallucination. It highlights research showing significant accuracy improvements (e.g., 16% to 72% in QA) and cost reductions (e.g., -25% tokens) when LLMs are grounded in ontologies and structured knowledge.

Key takeaway

For AI Architects and VPs of Engineering evaluating AI investments, recognize that current AI productivity claims are often unsupported by data. Your teams should prioritize building a foundational knowledge infrastructure, including ontologies and knowledge graphs, to achieve reliable, accurate, and cost-effective agentic AI systems. This requires shifting budget and cultural focus from solely data-centric approaches to structured knowledge work, treating it as critical, long-term infrastructure.

Key insights

Reliable agentic AI requires robust knowledge infrastructure, not just data, to overcome current productivity and hallucination issues.

Principles

• AI is a knowledge tool, not a data tool.
• Knowledge infrastructure must be built progressively, layer by layer.
• Structured knowledge significantly improves AI accuracy and efficiency.

Method

Build knowledge infrastructure sequentially: inventory assets, define controlled vocabularies, organize taxonomies, model ontologies, operationalize knowledge graphs, and establish continuous governance.

In practice

• Implement ontology-grounded RAG for higher accuracy.
• Define domain-specific vocabularies to reduce token usage.
• Utilize SPARQL queries for semantic entity reconciliation.

Topics

• Knowledge Graphs
• Agentic AI
• Ontologies
• Semantic Grounding
• LLM Accuracy

Code references

• nateburley/WikiKnowledgeGraphs

Best for: VP of Engineering/Data, AI Architect, AI Product Manager, CTO, Director of AI/ML, Executive

Related on AIssential

• Build Meaning Before Machines: Why Semantics, Ontologies, And Knowledge Graphs Matter For Agentic AI — Agentic AI requires explicit data meaning, making semantic layers and knowledge graphs essential architectural components.
• The Ontology Pipeline™, Refresh — Effective semantic knowledge infrastructure requires a structured pipeline, human expertise, and robust governance, not just AI-generated outputs.
• How to Build an Efficient Knowledge Base for AI Models — A systematic, iterative approach to knowledge base construction is essential for AI model accuracy and scalability.
• Knowledge Graph Engineering For Agents: Architecting For Self-Improvement — Agent reliability hinges on complete knowledge graphs; early deployment without robust knowledge leads to failure.
• The Context Gap — "Context graphs" are a market response to the self-inflicted wound of lost process knowledge due to decades of outsourcing.
• Agentic Architecture Part 3: The Information Layer — The Information Layer is foundational for agentic platforms, requiring incremental knowledge graph deployment for efficiency.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Intentional Arrangement.