How Samsara Built a Multi-Agent System to Drive a 36x Lift in Meeting Bookings on Databricks

Samsara faced a critical bottleneck: sales reps spent 60 minutes researching each prospect, yet meeting booking rates stagnated at 0.18%. To solve this, we built 'Smart Flows' on Databricks—a multi-agent system orchestrating Agent Bricks AI agents, AI/BI Genie and a custom MCP server for deep web research. This architecture moves beyond simple automation to agentic reasoning. A supervisor agent routes tasks: the MCP server gathers account intelligence, while Genie agents query CRM data to hyper-personalize outreach. The state is managed via Lakebase (Serverless Postgres) and reusable Unity Catalog functions. The results were transformative. Prospecting time dropped to under five minutes—12x productivity—and agentic personalization drove meeting booking rates to 6.49%—a massive 36x uplift. We will demonstrate this live, generating a fully researched outreach sequence in real-time to show how agents can drive hard revenue metrics.