Automate where repetition and risk intersect
Critical workflows, frequent regressions, and checks that consume the same manual effort every release are strong candidates. If a check is rare or constantly changing, manual coverage may still be the better trade.
Stability matters as much as importance
A highly important check can still be a bad automation target if the interface or behavior is too volatile. Sometimes the right move is to automate lower in the stack first.
Maintenance cost is part of the quality equation
Automation only pays off when it keeps paying off. If a test requires endless repair, its business value may be lower than the reporting suggests.
AI does not remove the prioritization problem
AI can help generate checks more quickly, but it does not decide which checks deserve long-term maintenance. Teams still need heuristics for value, risk, and cost.