United Express and Caylent build intelligent weight and balance on AWS

Because calculating the W&B of an aircraft is a critical step, the W&B system governs every departure, with roughly 530 regional jets operating across carrier partners. Delays are challenging and costly for customers, including the downstream implications of customer delays, impacted flights, and overall resources necessary for ground maintenance. A few years ago, United Express moved from a paper-based W&B process to a digital one. That transition was a meaningful step, but it still left engineers monitoring a high-volume system largely by hand.

The volume of data that the system produced made manual oversight unsustainable. Eight million log entries flowed in daily from multiple external sources, and engineers had no automated way to surface anomalies as they emerged. Core support engineers spent an estimated 336 hours per month on manual log analysis, time that was pulled directly from development and higher-value work. The team frequently reacted to issues after they had already affected operations. They needed to detect issues as they formed, engage the right people immediately, and build pattern recognition to stop events from recurring.

United Express worked with Caylent to apply the AWS Business Value Realization framework to evaluate potential use cases against business impact and speed to value. That structured assessment confirmed the W&B observability gap as the highest-priority opportunity. After validating the use case, the team designed and built an agentic AI observability solution for monitoring the flight system in NRT. At the center of the solution is Amazon CloudWatch, an intelligent observability service that provides actionable insights across applications and infrastructure. The service ingests logs from across the W&B system. When event, error, or performance thresholds are exceeded, Amazon CloudWatch routes alerts to the company’s dashboards and messaging applications, reaching the right engineers immediately. The solution also identifies emerging error patterns and event combinations for further investigation, shifting the team from reactive responses to proactive improvement.

For deeper analysis, the team built an AI agent on Amazon Bedrock, a service that gives access to hundreds of foundation models from leading AI companies. The solution collects data from events and API calls asynchronously and transforms it. Engineers can query that data through a conversational interface to ask questions in plain language and receive answers in seconds rather than hours. The solution connects directly to the company’s existing key-performance-indicator and application dashboards, giving stakeholders a unified view of system health.

Using the new solution, United Express reduced investigation time from roughly 10 hours to 1–2 hours. The company also cut mean time to detect and resolve issues by about 90 percent. Core support engineers now act on issues rather than searching for them, and business stakeholders receive updates faster. Event volume has fallen, cases close more quickly, and the time that once went to log analysis now goes to higher-value engineering work. Fast detection means that the company can resolve W&B issues before they can affect departures, reducing delay-related costs and limiting passenger disruptions.

Each resolved issue strengthens the team’s ability to catch similar issues earlier. “The quicker we can identify issues, the quicker we can fix or prevent them to deliver excellent customer service in the moments that matter most,” says Thomas Jacobs, product owner at United Express. “By implementing AI in our solutions, we’ve improved our operational efficiency and customer experience.” From a W&B system that once demanded hundreds of hours of manual scrutiny each month, United Express now operates with an AI-powered observability solution that catches issues in minutes. This helps protect departures, preserve schedules, and return engineering capacity to the work that keeps the airline network moving forward.