Databricks Agent Bricks Goes GA: 100K Agents, Multi-AI, and Key Summit Partnerships

Three core capabilities achieved General Availability (GA) earlier in 2026 and were highlighted at the summit:

• Custom Agents — Developers can use their preferred framework (LangGraph, PyFunc, and others) and deploy agents as fully managed Databricks Apps on serverless compute. This capability reached GA in February 2026 .
• Supervisor Agent — A multi-agent orchestrator that coordinates specialized sub-agents and tools for complex tasks. Supervisor Agent became GA on February 10, 2026, initially in select US regions .
• Document Intelligence — Processes unstructured data such as emails, PDFs, and manuals, moving beyond structured data alone .

These capabilities expanded regionally on Azure throughout early 2026. By March 31, Knowledge Assistant and Supervisor Agent were available in westeurope, northeurope, uksouth, centralindia, canadacentral, australiaeast, and southeastasia .

Databricks reported that more than 100,000 agents have been built on the platform since launch, processing over one quadrillion tokens per year of agent traffic .

Multi-Agent Orchestration and Multi-Model Flexibility

Two architectural themes emerged from the summit sessions and announcements.

Supervisor Agent: Coordinating Swarms

The Supervisor Agent pattern lets a central coordinator delegate tasks to sub-agents and tools. Databricks described this as moving from single-agent chatbots to coordinated multi-agent systems that can handle end-to-end business workflows . AstraZeneca's presentation provided the most detailed public example (see Enterprise Adoption below).

Any Model, One API

Agent Bricks is explicitly multi-model. Teams can use Anthropic Claude, Amazon Nova, Meta Llama (via Bedrock), and other providers through a single API with routing, fallbacks, and cost optimization . Databricks claims 63% of customers are already routing tasks across multiple model families .

Automated agent optimization, powered by Mosaic AI research, handles synthetic data generation, task-aware evaluation, and tuning, aiming to improve quality while reducing cost .

The Partnership Ecosystem at DAIS 2026

The summit also crystallized how third-party tools plug into the Agent Bricks lifecycle. Here is what was publicly confirmed:

Monte Carlo: Observability

Monte Carlo announced its integration with Agent Bricks on June 15, 2026, positioning itself as the exclusive observability partner for the platform . The integration extends Monte Carlo's existing Databricks coverage—from Delta Lake and Unity Catalog—into the agent layer, monitoring AI agent inputs, behavior, and outputs end to end .

Collibra: Governance

Databricks named Collibra its Data Governance Partner of the Year in the 2026 ISV Partner Awards . The deepened integration brings enterprise governance, semantic context, and AI lifecycle visibility directly into the Databricks Data Intelligence Platform, aiming to ground agentic AI in governed context with transparency and traceability .

Retool: Low-Code Agent Building

Retool was a launch partner for both Agent Bricks and Lakebase. Joint customers can build internal tools and custom agents using Retool's low-code platform connected to Databricks. A notable feature is the agent-to-agent connection with Genie Spaces, allowing teams to act on signals like revenue anomalies or inventory gaps directly in Retool .

Global Systems Integrators

Databricks' 2026 Global Partner Awards, announced during the summit, recognized major SIs:

• Accenture Federal Services won Global Partner of the Year .
• Deloitte took North America Partner of the Year and Public Sector SLED Partner of the Year .
• Wipro received Innovation Partner of the Year .
• Infosys was listed among Consulting and System Integrator Brickbuilder Award honorees .

Thoughtworks was not publicly named in DAIS 2026 partner awards, launch partner announcements, or Agent Bricks integration materials. We found no evidence of a specific Agent Bricks partnership announcement for Thoughtworks at this summit.

Enterprise Adoption: Who Is Running in Production

Databricks and its customers showcased production deployments across industries:

AstraZeneca: Scaling Multi-Agent Systems 10x

AstraZeneca presented a dedicated session titled "AstraZeneca's Multi-Agent System: Lessons Scaling Agents by 10x With Agent Bricks." The pharmaceutical company built the "Brand Assistant," which uses a Supervisor Agent to coordinate specialized sub-agents across therapeutic areas. The system combines structured data (via Genie Spaces) with unstructured documents (via Knowledge Assistant) and enforces permission boundaries through Unity Catalog integrated with Entra ID. The Model Context Protocol (MCP) enables third-party tool integration .

Virgin Atlantic

Virgin Atlantic was featured as a Databricks customer on Agent Bricks and presented at the summit . The airline's Head of BI and Advanced Analytics, Tom Barber, participated in sessions . Specific details of their agent deployment were available in summit on-demand content rather than public press releases.

Others

7-Eleven built a GenAI maintenance assistant on the platform . Fox Corporation, Flo Health, and EchoStar were also cited as production users .

Unconfirmed for This Summit

Workday had a summit session titled "Accelerating AI From Model to Impact at Workday With Databricks Apps," but it is not directly linked to Agent Bricks in available session descriptions . Zapier does not appear in relevant DAIS 2026 session listings or press materials for an Agent Bricks-specific announcement.

What This Signals

The 2026 Data + AI Summit marked a deliberate shift in messaging: from "here is what agents could do" to "here is what agents are already doing at scale." The combination of GA milestones, multi-agent orchestration, open model routing, and a maturing partner ecosystem suggests Databricks is positioning Agent Bricks as infrastructure, not a feature.

The unanswered questions—which will likely shape the next year of enterprise AI—center on governance at scale, agent-to-agent communication standards, and how observability and semantic context tooling from partners like Monte Carlo and Collibra actually perform under production loads. For now, the platform has moved firmly from prototype to production.