What challenges did Verizon Connect face before AI implementation?

Fleet managers were drowning in telematics data and had 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.

What was the impact of the AI solution?

Operational Insights was rolled out to Verizon Connect users in November 2025 and scaled to serve 100,000 users daily. The feature replaces manual analysis with clear, natural-language narratives and enables faster identification of safety, efficiency, and fleet performance patterns. The system supports cost-efficient delivery at scale by using concurrency controls and lower-cost model selection.

Source

Verizon Connect scales agentic AI for fleet anomaly insights to 100,000 users with Amazon Bedrock

Use case typeOperations optimizationUpdated May 27, 2026

Verizon Connect, a global fleet management solutions provider serving businesses worldwide through its Reveal platform, faced overwhelming telematics data volume that made it hard to identify emerging safety, maintenance, and operational inefficiency patterns from fragmented logs and spreadsheets. The company built a two-stage agentic AI solution on AWS that first detects anomalies with serverless orchestration and then uses parallel AI agents to investigate those anomalies and generate natural-language operational insights inside the Reveal application for fleet managers.

Organization: Verizon Connect
Industry: Logistics
Location: United States
Published: May 2026

Reported outcomes

+100%

harsh braking increase in example insightOther quantified impact

100,000 users/daydaily users served1,200,000 subscriptionsfleet subscriptions500,000,000 data points/daydata points generated daily−70%input token cost reduction with Amazon Nova 2 Lite+50%idling engine-on time in example insight

Strategic outcomes

Cost efficiencyReplaced manual analysis with automated operational insightsCustomer experience & trustDelivered natural-language narratives inside the Reveal applicationScale & capacitySupported fleet insight generation at enterprise scale

Primary read

Use case focus

Showing 2 of 2

1Operations optimization
2Safety monitoring

Fleet managers were drowning in telematics data and had 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.

Verizon Connect implemented a two-stage agentic architecture. A serverless statistical model orchestrated with AWS Step Functions detects anomalies from structured data and stores them for downstream analysis.
After anomalies are prepared, multiple AI agents run in parallel in AWS Lambda using the Strands Agents Framework. Those agents use Amazon Bedrock for language generation, query Amazon S3 for raw and anomaly data, use Amazon DynamoDB to track request status, and use Amazon SQS to manage concurrency for timely delivery.
The agents dynamically investigate anomalies with tool-based reasoning loops, pull fresh context at analysis time, and generate natural-language insights that are surfaced in the Reveal application.
The team optimized cost and throughput by validating with Claude 4.5 Sonnet, moving to Claude 4.5 Haiku, and then using Amazon Nova 2 Lite to reduce input token costs while maintaining quality through automated testing and a gold-standard dataset.

Technologies

Amazon Bedrock AWS Step Functions Amazon SQS Amazon S3 Amazon DynamoDB AWS Lambda Strands Agents Framework Amazon Nova Lite

Operational Insights was rolled out to Verizon Connect users in November 2025 and scaled to serve 100,000 users daily.
The feature replaces manual analysis with clear, natural-language narratives and enables faster identification of safety, efficiency, and fleet performance patterns.
The system supports cost-efficient delivery at scale by using concurrency controls and lower-cost model selection.

Architecture

A serverless anomaly detection stage is orchestrated with AWS Step Functions and stores results, then parallel AI agents run in AWS Lambda via the Strands Agents Framework. The agents retrieve anomalies from Amazon S3, query raw and historical context, write insights back to Amazon S3, track status in Amazon DynamoDB, and use Amazon SQS to control concurrency and stay within Amazon Bedrock quotas.

Sources & evidence1

Groundedness: 5/5Type: Blog PostPublished: May 27, 2026Publisher: AWSEvidence: VendorConfidence: Medium

Primary source

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases

Customer evidence