Run acceptance tests as soon as code for a story or a testable component of a story is finished. Keep running the tests every build.
Run acceptance tests from previous iterations as regression tests for each build in the development integration environment.
Ideally, the customer runs the tests, but when he doesn't participate fully on a daily basis, programmers and testers run the tests until the customer has a chance to do so.
Acceptance tests (except for those dependent on other uncompleted stories) must pass before the stories are complete. If the customer can't run them until after the end of an iteration, for whatever reason, the programmers must run them.
Educate the customer before and during the project so she understands the testing, defect-reporting, and defect-fixing processes.
The customer defined quality criteria at the start of the iteration. Verify that these criteria were met.
Outstanding defects should be written up as stories, estimated, and put in the bucket of stories from which the customer can choose for the next iteration.
