Verizon Connect scaled agentic AI solution to serve 100,000 fleet managers daily using two-stage reasoning architecture

Verizon Connect, a fleet management solutions provider, deployed an agentic AI system (autonomous AI that decides its own investigation steps) to transform data from over 1.2 million active vehicle subscriptions generating over 500 million data points daily into actionable insights for 100,000 users daily.

The system uses a two-stage approach: Stage 1 aggregates detected anomalies into insight candidates and scores them by severity and impact; Stage 2 spawns separate agent instances that autonomously call data retrieval tools to investigate each insight, adapting their investigation path based on what the data reveals rather than following fixed rules.

By offloading numerical analysis to specialized statistical code rather than asking the LLM to analyze raw tabular data directly, the architecture avoids the scale and accuracy issues LLMs face with large structured datasets, allowing the AI agent to focus on synthesizing the 'why' and 'how to address' dimensions of each anomaly.