Transportation Management Platform — ML Support Ticket Classification

10,000 messages a day. 91% of the noise eliminated. Support agents doing support again — not sorting.

A transportation management company was losing agent time to irrelevant messages flooding the support queue, promotions, general enquiries, and non-support content mixed with urgent issues. Manual triage was slowing response times and impacting customer satisfaction. VeUP built a real-time ML classification pipeline using a fine-tuned RoBERTa transformer on Amazon SageMaker, with LLM-assisted labelling and continuous drift detection.

In production

• 91% reduction in irrelevant tickets entering the support queue
• 60% increase in agent productivity and response capacity
• 10,000+ messages classified per day with consistent accuracy
• Real-time and batch processing both live
• Model Monitor providing continuous drift detection and auditability

AWS Services:

Amazon SageMaker

Amazon S3

Amazon Athena

Amazon SageMaker Model Monitor