The post will show how the multi-agent supervisor pattern accelerates production-ready Generative AI deployment while maintaining the specialized knowledge.

databricks

BASF Coatings partnered with Databricks to implement a production-ready multi-agent AI system using a supervisor architecture pattern. The solution, called Marketmind, orchestrates specialized agents including Genie for structured data queries and function-calling agents for unstructured data retrieval through vector search. Built on Databricks' Mosaic AI framework and integrated with Microsoft Teams, the system consolidates Salesforce customer data, market insights, and external trends to provide sales teams with actionable intelligence. After a 5-6 week proof of concept and one-month pilot with 25 users, the system is rolling out to over 1,000 sales representatives globally, with plans to expand across Supply Chain, Procurement, and other divisions using a hierarchical supervisor-of-supervisors architecture.

Multi-Agent Supervisor Architecture: Orchestrating Enterprise AI at Scale

Looking Forward: Multi-Layered Orchestration and Agent Bricks