This paper presents the intelligent automation of functional test of services (unit testing) and services architectures (end-to-end testing) that has been developed by the MIDAS project, and is accessible on the MIDAS SaaS. In particular, the paper illustrates how the extreme automation implemented in the MIDAS prototype includes the solutions of tough problems such as: (i) the configuration of the automated test execution engine against large and complex services architectures, (ii) the test input generation based on formal methods and constraint propagation, (iii) the test oracle generation based on state machine execution, (iv) the dynamic scheduling of test cases and the reactive, evidence-based planning of test campaigns with on the fly generation of new test cases, both based on based on probabilistic graphical inference. This paper reports some feedback from real-world case studies in the health-care and in the logistics sectors.