One thing that was not asked by @DaAnalyst specifically but is of value for the topic but use to be hard to do and now much easier with AI help is to update test coverage to 100% on the core code, not 98%, not 99%. The reason being that if the AI has access to the unit test, it can use that as the source of truth, and also learn more about how the code works and is expected to work rather than spending hours of reasoning on its own coming to false beliefs.
This reasoning is very much in line with the evidence reply from another topic.
