Computational Intelligence has achieved considerable successes in recent years and numerous applications rely on it. However, its use crucially relies on human experts, from feature engineering to algorithm selection. Since the performance of a given approach depends on the qualities of the algorithm regarding the targeted data and task, the design of new application becomes more and more difficult. Benchmarking, algorithm selection and hyperparameter tuning thus become tools of choice for helping non-experts to rigorously design the method that will solve their problem. This presentation aims at introducing the (relatively) new scientific field arising around the progressive automation of computational intelligence.