OFICIAL Databricks Newsroom

How to scale agentic applications without creating AI sprawl

What happened
Based on Databricks Newsroom · Sep 30, 2026

Enterprises face AI sprawl as agentic applications multiply, requiring shared infrastructure for choice, context, and control to scale without duplicating integrations or governance.

How to scale agentic applications without creating AI sprawl
Databricks Newsroom — Databricks
Key points
·
Agent Bricks, Omnigent, and Unity Gateway provide shared infrastructure for building, governing, and scaling agent fleets without duplicating integrations.
·
Unity Catalog and Genie Ontology enable governed reuse of enterprise data and business semantics across agentic applications.
·
Omnigent and Unity Gateway allow teams to switch models and harnesses while maintaining consistent access, routing, and cost controls.

As agents evolve from answering questions to executing actions across systems, enterprises risk fragmented integrations, inconsistent policies, and rising costs without a unified foundation. A single workflow may involve retrieving governed data, selecting models, invoking tools, and updating business systems while maintaining permissions and traceability. Without shared infrastructure, each team’s independent wiring of these components can lead to duplicated efforts and governance gaps as the number of agents grows.

Databricks, OpenAI, and Stellantis recently collaborated to address scaling agentic applications in production, emphasizing the need for infrastructure that provides choice, context, and control. Agent Bricks offers a unified platform for building and governing agent fleets, while Omnigent serves as a common layer across agent harnesses and Unity Gateway centralizes access, cost controls, and observability. These shared capabilities allow teams to add new agents without recreating integrations or governance structures for each one.

Agentic applications require more than model access; they need governed enterprise data, business definitions, and tools to understand tasks accurately. Without a shared context layer, concepts like customer definitions or product ownership may vary across workflows, creating maintenance burdens. Databricks integrates Unity Catalog for data governance and the Genie Ontology for business semantics, enabling teams to reuse governed context across applications instead of rebuilding it repeatedly.

Flexibility in model and harness selection is critical as workflows evolve, but managing diverse interfaces and policies becomes complex. Omnigent provides a common layer above agent harnesses, allowing teams to compose agents with different technologies and switch between them with minimal rework. Unity Gateway offers consistent access to models, Smart Routing, capacity, and cost controls, ensuring the surrounding infrastructure remains stable even as underlying components change.

Original source → Deals on Clipraptor.com →