Software process models formalize the way a group of agents (e.g. developers, testers, managers etc) interact in order to produce a desired outcome (e.g. a product, an artifact etc). In this context, a "deviation" is a mismatch between the process executed by the agents and the process model. Existing approaches for deviation detection and handling force the agents to either pursue a deviation-free process execution, which is unrealistic; or to selectively ignore them, which may be risky to the desired outcome of the project. In this paper, we propose an approach that allows agents to deviate from the process specification, but also allows them to correct these deviations later in the process enactment. Additionally, they are informed about the risks implied by each non-handled deviation. During the correction phase, the process agents are assisted by the means of a set of correction plans that are automatically generated by the approach. These plans aim at reducing the risk of non resolved deviations. This paper presents a preliminary evaluation of this approach as a prototype implementation.