Semantic Foundations for Reliable Enterprise AI

2026-06-19 · Source: Modern Data 101 · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Software Development & Engineering · Depth: Intermediate, medium

Summary

The article by Anvar Atash addresses why enterprise large language models (LLMs) frequently generate inaccurate financial figures, attributing the issue to semantic data mismatches rather than model architecture or retrieval-augmented generation (RAG) pipelines. It highlights that schema changes in ML/LLM pipelines often lead to undetected quality deterioration, unlike traditional dashboards where errors are immediately apparent. The proposed solution emphasizes robust data contracts that define not only schema but also underlying semantics. It introduces the "Ontology Pipeline," a systematic, layered framework by Jessica Talisman, for constructing semantic knowledge management systems that culminate in a knowledge graph. This approach enables LLMs to achieve precise understanding, moving beyond mere inference to deliver accurate, context-specific information, such as net revenue calculated per IFRS definition. The piece also acknowledges significant cultural challenges in implementing data contracts, recommending a phased approach focusing on high-impact areas like financial data.

Key takeaway

For AI Architects and Data Engineers building enterprise AI systems, recognize that semantic data consistency is paramount for reliable LLM outputs. Your focus should shift from solely optimizing RAG pipelines to implementing comprehensive data contracts that explicitly define data semantics. Prioritize adopting the Ontology Pipeline to build a robust knowledge graph, ensuring your AI models deliver precise, contextually accurate information rather than misleading inferences. This proactive approach mitigates undetected data quality issues.

Key insights

Semantic clarity in data, achieved through robust data contracts and ontologies, is crucial for reliable enterprise AI outputs.

Principles

• LLM performance is constrained by inconsistent data, not model architecture.
• Data contracts must define semantics, not just schema.
• Undetected data quality issues degrade AI model outputs.

Method

The Ontology Pipeline, an iterative framework by Jessica Talisman, builds semantic knowledge management systems from controlled vocabularies to knowledge graphs.

In practice

• Prioritize data contracts for high-impact financial data.
• Implement schema registries to track contract modifications.
• Ground LLMs in knowledge graphs for precise understanding.

Topics

• Enterprise AI
• Data Contracts
• Semantic Data Management
• Ontology Pipeline
• Knowledge Graphs
• LLM Reliability

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

Related on AIssential

• The Ontology Engineering Challenge is Not Building It, but Mirroring the Delta — The core challenge in enterprise AI is correcting LLM's inherent concept understanding where it diverges from specific business definitions.
• The Ontology Pipeline™, Refresh — Effective semantic knowledge infrastructure requires a structured pipeline, human expertise, and robust governance, not just AI-generated outputs.
• Semantic Layers for Reliable LLM-Powered Data Analytics: A Paired Benchmark of Accuracy and Hallucination Across Three Frontier Models — Explicit business semantics significantly improve LLM accuracy and reduce hallucination in natural-language data querying.
• The Semantic Layer and AI Agents // David Jayatillake // MLOps Podcast #343 — A semantic layer standardizes data definitions, enabling AI agents to automate data tasks and answer business questions efficiently.
• Ontologies Are the Intentional Core of a True Knowledge Graph — Formal ontologies provide the stable, intentional semantic backbone essential for reliable enterprise AI and knowledge graphs.
• Steal This Deck — Reliable agentic AI requires robust knowledge infrastructure, not just data, to overcome current productivity and hallucination issues.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Modern Data 101.