Imagine each task you might possibly do for a customer (the responses to your post suggest many). At some point, 0% of the task is automated. You can do it for customers 100% manually.
If you do this for all of the tasks people will pay you to do, you can get a sense of "which ones are slow, time sensitive, easy to automate, or frequent"
Once you have that, you can now prioritize work that will make those tasks more efficient. If you just thought about tasks in a vacuum, you wouldn't have any empirical data that tells you what is worth doing. By doing things that don't scale - by doing things completely manually - you've also figured out a roadmap for the next few days/weeks.
If you do this for all of the tasks people will pay you to do, you can get a sense of "which ones are slow, time sensitive, easy to automate, or frequent"
Once you have that, you can now prioritize work that will make those tasks more efficient. If you just thought about tasks in a vacuum, you wouldn't have any empirical data that tells you what is worth doing. By doing things that don't scale - by doing things completely manually - you've also figured out a roadmap for the next few days/weeks.