An AI modeling company experiences unpredictable traffic surges during research simulations. They deploy large Amazon EC2 instances from a custom Amazon Machine Image (AMI) using an Auto Scaling group. To maintain performance during peak demand, the company needs a way to launch new instances quickly with the least possible delay during initialization. What approach should a solutions architect take to reduce the launch time of new instances and meet the responsiveness requirement?