Agentic data stack: From pipelines to semantic intelligence

Enterprise data analytics stands at an inflection point. Traditional deterministic pipelines demand endless maintenance cycles, where engineering teams spend more time on data cleaning than on analytical innovation. This talk examines how large language models and large reasoning models reshape this landscape through semantic intelligence.

The agentic data stack introduces three foundational pillars transforming analytics infrastructure. First, standardized context protocols, particularly the Model Context Protocol, establish universal interfaces between reasoning models and execution environments, decoupling intelligence from tooling. Second, semantic intelligence layers embed business logic and metadata directly into storage systems, enabling systems to interpret intent autonomously and reason about schema compatibility without explicit programming. Third, unified cognitive architectures virtualize analytics across disparate tools and cloud platforms, unlocking zero-ETL patterns and comprehensive data lineage visibility.

Audience members will discover:
• How agents autonomously handle legacy code modernization and text-to-SQL accuracy challenges
• The economics of GenAI-powered migration agents compared with traditional labor-intensive approaches
• Implementation patterns for self-healing pipelines and predictive optimization systems
• Governance frameworks preventing Model Autophagy Disorder while maintaining human-AI collaboration

Organizations adopting these solutions transition from syntax-focused engineering to semantic-driven analytics, compressing skill gaps while freeing capacity for innovation. This session provides actionable patterns for platform teams architecting the next generation of data infrastructure.