SourceInfoQ中国•Stalecollected in 0m
Snowflake Cortex Agents Use Ontology-Driven Reasoning
#ontology#enterprise-agents#knowledge-groundingsnowflake-cortex-agentssnowflakesnowflake-cortex-agents
Learn how ontology design can help enterprise agents reason with business context.
30-Second TL;DR
What Changed
Applies ontology-driven reasoning to Snowflake Cortex Agents
Why It Matters
Ontology-based grounding can make enterprise agents more consistent and aligned with domain terminology. It may also improve the reliability of analytics and workflow automation where business context is essential.
What To Do Next
Prototype a small business ontology for one analytics domain and evaluate Snowflake Cortex Agents on terminology accuracy and relationship-based questions.
Who should care:Enterprise & Security Teams
Key Points
- •Applies ontology-driven reasoning to Snowflake Cortex Agents
- •Helps agents interpret business concepts, entities, and relationships
- •Presents a practical framework for grounding agent responses in business knowledge
Deep Insight
AI-generated analysis for this event — not the original article.
Enhanced Key Takeaways
- •Snowflake Cortex Agents leverage the 'Cortex Analyst' framework to bridge the gap between natural language queries and structured SQL execution by utilizing semantic models.
- •The ontology-driven approach utilizes YAML-based semantic models that define business metrics, dimensions, and relationships, preventing the 'hallucination' of non-existent database columns.
- •Integration with Snowflake's 'Document AI' allows these agents to extract unstructured data from documents and map them directly into the defined business ontology.
- •The architecture employs a multi-step reasoning process where the agent first validates the user's intent against the ontology before generating the corresponding query plan.
- •Snowflake has optimized these agents to run entirely within the Snowflake security perimeter, ensuring that sensitive business logic and data do not leave the customer's governance boundary.
Competitor Analysis
Core Approach
- Snowflake Cortex Agents
- Ontology/Semantic Model-driven
- Databricks AI/BI Agents
- Unity Catalog/Lakehouse-driven
- Microsoft Fabric Copilot
- OneLake/Semantic Link-driven
Governance
- Snowflake Cortex Agents
- Native Snowflake RBAC
- Databricks AI/BI Agents
- Unity Catalog
- Microsoft Fabric Copilot
- Microsoft Purview
Primary Focus
- Snowflake Cortex Agents
- SQL/Structured Data Reasoning
- Databricks AI/BI Agents
- Data Intelligence/Notebooks
- Microsoft Fabric Copilot
- Power BI/Office Integration
| Feature | Snowflake Cortex Agents | Databricks AI/BI Agents | Microsoft Fabric Copilot |
|---|---|---|---|
| Core Approach | Ontology/Semantic Model-driven | Unity Catalog/Lakehouse-driven | OneLake/Semantic Link-driven |
| Governance | Native Snowflake RBAC | Unity Catalog | Microsoft Purview |
| Primary Focus | SQL/Structured Data Reasoning | Data Intelligence/Notebooks | Power BI/Office Integration |
Technical Deep Dive
- Uses a YAML-based semantic model to define the business ontology, mapping natural language terms to specific tables and columns.
- Implements a RAG (Retrieval-Augmented Generation) pipeline that retrieves relevant schema metadata before LLM inference to ensure query accuracy.
- Employs a 'Plan-and-Execute' agentic pattern where the model generates a SQL plan, validates it against the semantic model, and then executes it.
- Supports multi-turn conversation state management, allowing the agent to maintain context of previous filters and business definitions during a session.
- Utilizes Snowflake's serverless compute infrastructure to scale inference tasks dynamically based on query complexity.
Future ImplicationsAI analysis grounded in cited sources
Autonomous data engineering will become the standard for enterprise BI.
By grounding agents in ontologies, organizations will shift from manual report creation to self-service, agent-led data discovery.
Semantic modeling will replace traditional metadata management tools.
The requirement for LLMs to understand business context necessitates a centralized, machine-readable ontology that serves as the single source of truth.
Timeline
2023-11
Snowflake announces Cortex, a managed AI service for LLMs and vector search.
2024-04
Snowflake introduces Cortex Analyst in private preview to enable natural language interaction with data.
2024-06
Snowflake announces Document AI integration to expand Cortex capabilities to unstructured data.
2025-02
Snowflake expands Cortex Agent capabilities to support complex, multi-step reasoning workflows.
- 2023-11Snowflake announces Cortex, a managed AI service for LLMs and vector search.
- 2024-04Snowflake introduces Cortex Analyst in private preview to enable natural language interaction with data.
- 2024-06Snowflake announces Document AI integration to expand Cortex capabilities to unstructured data.
- 2025-02Snowflake expands Cortex Agent capabilities to support complex, multi-step reasoning workflows.
Weekly AI Recap
Read this week's curated digest of top AI events →
AI-curated news aggregator. All content rights belong to original publishers.
Original source: InfoQ中国 ↗
This is a summary, not the original. Read the source, or get the weekly briefing.
The weekly digest
One email a week. Unsubscribe anytime.