Workflow management systems usually interpret a workflow definition rigidly. However, there are real life situations where users should be allowed to deviate from the prescribed static workflow definition for various reasons, including lack of information and unavailability of the required resources. To flexibilize workflow execution, this paper proposes mechanisms that allow execution to proceed in the presence of incomplete information, by adopting presuppositions, and in the presence of negative information, by suggesting execution alternatives. This paper also presents an architecture for the workflow system, which is driven by ontologies that capture semantic relationships between workflows, resources and users.