A B2B e-commerce company wants a simple, operationally light approach to reject fraudulent transactions, accepting some loss of profitable transactions or customers. Which solution will meet this need with the least operational effort?

The correct answer is to use the Amazon Fraud Detector prediction API to automatically approve or deny any activity that Fraud Detector flags as fraudulent. This approach directly addresses the requirement for a simple, operationally light solution that automatically rejects fraudulent transactions. Amazon Fraud Detector is a fully managed service designed for this purpose, requiring minimal operational overhead once configured. Using Amazon SageMaker to only approve transactions for products previously sold is too restrictive and would lead to a high number of false positives, rejecting legitimate new business. Training a custom fraud-detection model in Amazon SageMaker is a more complex and operationally intensive solution, contrary to the "least operational effort" requirement. Using the Amazon Fraud Detector prediction API to flag potentially fraudulent activities for human review adds a manual step, increasing operational effort and not fully meeting the "automatically reject" aspect of the requirement.