CPU utilization is a primary indicator of how much work a system is performing and is commonly used to trigger scaling actions. When the demand increases, resulting in higher CPU usage, an Auto Scaling group can automatically launch additional EC2 instances to distribute the load. Conversely, it can terminate instances during periods of low CPU usage to reduce costs. Network I/O may also indicate increased demand, but it's not as common a trigger for scaling as CPU utilization. The number of users is not a direct metric used by Auto Scaling. Disk Read/Writes can indicate IO intensive workloads, but they are not a primary scaling metric in comparison to CPU utilization.