The current paper describes an attempt to introduce data-driven learning of Greek as mother tongue in the context of primary and secondary education. The above action was implemented for the Greek Ministry of Education and Religious Affairs and involved the construction of a pedagogically relevant corpus and the design of pedagogical scenarios of use. The corpus (Educational Thesaurus of Greek Texts) comprises a total of about 43 million tokens distributed in three sub-corpora: a reference corpus which is part of the Hellenic National Corpus (HNC), a pedagogical corpus with texts from the school textbooks and an open teacher-compiled corpus. The educational scenarios cover a total of 150 teaching hours and suggest a series of corpus-based learning activities primarily for the Greek language lesson but also for other subjects in the school curriculum. The paper also presents indicative results from the preliminary small scale evaluation of the Educational Thesaurus of Greek Texts by 85 pupils. The main advantages of data-driven learning are discussed, such as the enhancement of autonomous and discovery learning. Finally, the paper discusses some of the main obstacles in introducing corpora in the language classroom. These are relevant both to the nature of the corpus itself and to the limitations inherent to the strict curriculum that the educational scenarios should be in line to.