Software testing, in particular acceptance testing, is a very important step in the development process of any application since it represents a way of matching the users' expectations with the finished product´s capabilities. Typically considered as a cumbersome activity, many efforts have been made to alleviate the burden of writing tests by, for instance, trying to generate them automatically. However, testing still remains a largely neglected step. In this paper we propose taking advantage of existing requirement artifacts to semi-automatically generate acceptance tests. This paper extends a previous paper in which we use Scenarios, a requirement artifact used to describe business processes and requirements, and Task/Method models, a modelling approach taken from the Artificial Intelligence field. The proposed approach derives a Task/Method model from Scenario (through rules) and from the Task/Method model specification, all alternatives in the flow of execution are provided. Using the proposed ideas, we show how the semi-automated generation of acceptance tests can be implemented by describing an ongoing development of a proof of concept web application designed to support the full process.