“A strategy mentioned for asking the agent what could have been done better at the end of a run to identify errors.”