Automating KPI Reasoning and Narrative Generation With Agent Bricks and Databricks Apps

Monthly performance reporting previously required approximately 7 person-days each month, as analysts manually wrote queries, reviewed distributed datasets, interpreted metric relationships and produced presentation-ready reports. This repetitive process tied up analytical resources and slowed decision-making. To address this, I built an app-based automation pipeline using multi-supervised agents with Agent Bricks and Databricks Apps. The agent operates as an analyst, systematically analyzing relevant data sources through a structured, table-driven representation to identify key drivers, evaluate metric relationships, and generate narrative insights without manual querying. Outputs are delivered through Databricks Apps as a self-serve analytics platform with automatically refreshed metrics and visualizations, eliminating repetitive reporting work, reducing monthly effort by approximately 7 person-days, and enabling agent-authored reporting for both analytics and non-analytics stakeholders.