Detecting Large-Scale Data Dashboard Errors with Amazon Bedrock
An AWS team has developed a new solution using Amazon Bedrock and serverless services to automatically detect silent visual and numerical errors in data dashboards.
A new automation solution has been implemented to prevent silent data dashboard errors—frequently encountered by organizations running business intelligence at scale and left undetected by infrastructure monitors. Thanks to this developed system, error detection time has been significantly reduced.
Silent Errors in Data Dashboards
In business intelligence processes, even if the infrastructure is operating completely healthily, users may encounter blank charts or incorrect data on screens. Such issues are generally not noticed by infrastructure monitoring systems.
Automated Content Validation Solution
The AWS team built a final-stage automated content validation solution that simultaneously scans hundreds of dashboards and detects missing or incorrect elements. This system enables the analytics team to resolve issues rapidly.
The Role of Large Language Models
The developed solution visually analyzes dashboards using large language models on Amazon Bedrock. When the system detects a health issue, it alerts the creators in real time.
Performance Results Achieved
Thanks to the new solution, the average detection time has dropped from periods reaching up to 72 hours to under 1 hour. Over a 30-day production operation, thousands of automated checks were performed, and the system ensured high content availability.