This paper is concerned with the adaptive execution of workflows on resources consisting of a pool of machines and a pool of alternative web services. The hierarchical nature of workflows enables adaptation at multiple levels. In this work, adaptivity is concerned with changing the mapping of services to machines and workflow invocations to services, in order to meet the requirements of both user and provider. Specifically, a third-party workflow engine (ActiveBPEL) has been wrapped to supportmapping at these two levels. Results of experiments using the prototype within a cluster environment are presented which demon- strate a benefit from adapting in response to changes of user load and to changes in the pool of alternative services available during a workload. The experiments include a range of adaptivity scenarios and show that, by selection of an appropriate policy, a significant gain can be made.