DAIS 2026: From Observability to Agentic Data Engineering

Data observability is now standard in lakehouse platforms, but it stops at reactive monitoring. As teams push toward proactive agentic workflows, LLMs alone aren’t enough. Without production-aware context, ability to take action in-flight, and automated guardrails, AI remains limited to suggestive copilots trained on generic data.

In this session from Databricks Data + AI Summit 2026, Roy Daniel breaks down what it takes to build agentic capabilities to continuously optimize cost, resolve incidents in-motion, and evolve pipeline code.Roy shows how definity enables them through full-stack context across data, pipelines, and infrastructure; runtime execution control; and built-in validation loops — turning signals into autonomous action inside Spark pipelines.