Unlocking semantics for AI: How Mercedes-Benz Korea built trusted “Talk to Data” at scale

2026-06-11 · Source: Databricks · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cloud Computing & IT Infrastructure · Depth: Advanced, long

Summary

Mercedes-Benz Korea collaborated with Databricks to implement a "Talk to Data" capability, establishing a unified, AI-ready, and governed semantic foundation for enterprise decision-making. This initiative extended their existing analytics infrastructure, which already managed over 500 KPIs across various business domains, by integrating Unity Catalog Business Semantics, Metric Views, Genie, and Agent Bricks on the Databricks Data Intelligence Platform. A key development was an automated DAX-to-Metric-View transpiler, which efficiently converted Power BI DAX measures into Databricks metric views, significantly reducing manual effort. The project also defined a five-phase process for curating AI-ready semantics, targeting 100% answer alignment with Power BI reports, and deployed a multi-agent system with persona-based access control. This pilot provides a repeatable playbook for other Mercedes-Benz markets to adopt a consistent, explainable, and high-quality AI-driven data interaction experience.

Key takeaway

For AI Architects or Directors of AI/ML building enterprise "Talk to Data" solutions, you should prioritize establishing a unified, governed semantic layer. This approach ensures consistent KPI definitions and business logic across both BI and AI, directly improving answer reliability and explainability. Implement a DAX-to-Metric-View transpiler to accelerate migration of existing BI semantics. Additionally, deploy persona-based agents with granular access control via Unity Catalog to tailor user experiences while maintaining data governance.

Key insights

Unified, governed semantic layers are crucial for reliable, scalable enterprise "Talk to Data" AI, bridging BI and AI definitions.

Principles

• Answer reliability requires governed business logic.
• Unified semantics support consistent BI and AI.
• Persona-based agents need tailored governance.

Method

A five-phase process (Prepare, Build, Organize, Test, Validate) ensures AI-ready semantics and 100% answer alignment with BI reports, iteratively refining metric views and Genie spaces.

In practice

• Transpile DAX measures to metric views.
• Limit Genie spaces to 30 Unity Catalog items.
• Use benchmarks for answer accuracy.

Topics

• Talk to Data
• Semantic Layer
• Databricks Unity Catalog
• Agentic AI
• KPI Governance
• BI-AI Integration

Best for: AI Architect, Director of AI/ML, Data Engineer

Related on AIssential

• Logical First, Physical Second: A Pragmatic Path to Trusted Data — Effective data architecture prioritizes business meaning and shared semantic models over immediate physical schema design.
• Scaling Enterprise Conversational Intelligence: Cross-industry Technology and Functional Solutions Powered by Databricks Genie — Databricks Genie enables natural language interaction with enterprise data, powering diverse cross-functional AI solutions.
• Why agentic analytics starts with a well-governed data layer — AI amplifies the need for robust data semantics and governance to ensure trusted, scalable analytical outcomes.
• Semantic Layers in the Age of AI are 100% Needed — Semantic layers translate natural language queries into data operations, ensuring consistent data interpretation for AI and users.
• Building AI-ready data: Vanguard’s Virtual Analyst journey — AI-ready data infrastructure, not just advanced models, is crucial for successful enterprise conversational AI deployment.
• The Memory of an Enterprise: 5 Surprising Truths About How Data Becomes Intelligence — An Enterprise Data Warehouse creates corporate memory, transforming raw data into actionable intelligence through structured processes.

Open in AIssential →

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