I would consider the following points to help me decide if a test should be automated:

How often does the test need to be executed? i.e. is that going to be a regression test? Sometimes the test will need to be executed once, but with a large set of data

How much time does automating this test will save me so that I can use my time in exploratory testing

How important is the test to the business; i.e. is the test scenario a typical user journey through the application

How complex is it to automate the test and how likely is it that the complexity doesn’t cause many false positives which increases results analysis time?

How likely is it that this test catches a defect?

How likely is it that a feature or functionality will break and what is the impact of it to the business? If it is high impact, then it should be automated to ensure it passes from release to release