While you and your pair partner are working on a task, you run unit tests to make sure you are not introducing side effect defects. As a system evolves, it may become impractical to run all the unit tests for the system for every change made. In this case, you may choose a subset of the unit tests to run after every code change. When a task is completed or you feel that you have a piece you can integrate or you are uncertain for any reason, run all of the unit tests for the system.

If all of the unit test in the system passed, check in your code and produce a system build.