Solution
Baylor's Enrollment Management team built an agent workflow on Agent Bricks (Mosaic AI) that pulls call recordings and metadata from the contact-center phone system into a medallion architecture on Databricks. A Knowledge Assistant grounded in Baylor's policies and procedures gives representatives in-call guidance, while a Multi-Agent Supervisor scores every call against standard operating procedures, surfaces sentiment trends and Q&A patterns, and links findings to the original transcript with timestamps and source citations. Unity Catalog provides FERPA-aware governance, with row-level security tied to Active Directory groups so different departments see only the calls relevant to their work. Supervisors query call data on demand in natural language through Databricks Genie and generate daily summaries in about two minutes.
Data flow
Baylor's phone system feeds call recordings and metadata via API into a medallion architecture on Databricks. Agent Bricks runs two agents over the curated calls — a Knowledge Assistant grounded in policies and procedures, and a Multi-Agent Supervisor that evaluates each call against standard operating procedures and links findings back to the original transcript with timestamps and source citations. Supervisors explore results through Databricks Genie under FERPA-aware Unity Catalog row-level security backed by Active Directory groups, and generate daily reports summarizing volume, sentiment, and coaching examples in a couple of minutes.