A CloudOps engineer is optimizing costs for an EC2-based e-commerce fleet. The application runs behind an Application Load Balancer, and CloudWatch shows average CPU utilization consistently between 8% and 15% across all instances during a full week of monitoring, including the daily peak traffic window. The instances are memory-intensive m5.2xlarge types, but the CloudWatch agent reports memory utilization never exceeds 30%. Network throughput and EBS metrics are also well below limits. What is the MOST appropriate action to reduce cost while maintaining performance?