When Should a Test Be Automated?

When the cost, measured in terms of missed bugs, of not automating exceeds the cost of bugs missed by tests never run even once because the time was devoted to automation.

See "When Should a Test Be Automated? ".

Automation protects against the case where a test was run (manually) once, the code was changed to inadvertently add a bug, and the test isn't then rerun to catch it because rerunning all manual tests is too expensive.

However, automation means fewer unique tests will be run, so there will be bugs that are not caught because no one checked for them even once.