[Databricks Data + AI Summit 2026] Analytics Platform for the AI Agent Era: Databricks Agentic Analytics Strategy

As generative AI evolves beyond simple answers into the Agentic AI era where it actually performs work, the role of data platforms is also changing. Databricks identified accuracy, governance, and scalability challenges as the core tasks for Agentic Analytics to boost the productivity of data teams.

Agents find it difficult to fully understand business context, and when various organizations create agents individually, Shadow IT phenomena can occur where security control becomes difficult. In addition, the problem of platform load that occurs when numerous agents query data simultaneously is also a challenge that must be solved.

To address this, Databricks proposes the following 4-layer architecture:

• Compute (Lakehouse RT): An ultra-high-performance engine that maintains response times of around 150ms even when thousands of agents query data simultaneously
• Governance (Unity AI Gateway): A gateway that centrally manages access to models, tools, and data
• Semantics: A layer that conveys business context to agents
• Agent Layer: The agent layer that actually performs users' work

Databricks emphasizes that the core of the agent era lies not simply in model performance, but in how fast and safely data can be provided.