With over 1.2 million active vehicle subscriptions generating over 500 million data points daily across 80,000 unique data indicators, fleet managers were drowning in this data and forced to hunt for anomalies across fragmented paper logs and reactive spreadsheets. The sheer volume made it impossible to identify emerging safety issues, maintenance needs, or operational inefficiencies before they became costly problems.Rather than building another static dashboard or rule-based automation system, which only catches predefined patterns, Verizon Connect chose agentic AI to replace that manual guesswork with a centralized intelligence solution. Agentic AI dynamically investigates new patterns, asks follow-up questions, and adapts its analysis based on what it discovers, making it well suited for the unpredictable nature of fleet operations.

In this post, we show you how Verizon Connect built and scaled an agentic AI solution to transform overwhelming fleet data into clear, actionable insights for 100,000 users daily. We walk you through the architectural decisions, implementation challenges, and measurable results that can guide your own data-to-insights transformation.

The solution handles data at scale while maintaining cost-efficiency. The following figure describes the core components. Later in this section, we walk through and discuss the various components of the solution and tie them together in the ‘Overall architecture’ section.

A common pitfall in AI engineering is asking an LLM to perform numerical analysis on large-scale raw tabular data. As AWS Prescriptive Guidance notes, LLMs can struggle with complex table structures and numerical extraction at scale. To address this, we built a serverless statistical model using AWS Step Functions and AWS Lambda (See Figure 4). This model performs the computationally intensive work of anomaly detection on structured data. It identifies what the anomaly is, so the AI agent can focus on why it occurred and how to address it.