The deluge of massive amounts of data (big data), poses questions of contemporary journalism, its practices, but also about the economic models of the media and the organization of editorial teams. The issues are multifaceted. First, we emphasize the epistemological reluctance of the profession to the promise of totally automated journalism. In fact, data journalism is always linked with an identified ground to which data must refer in order to validate it. Then we show that the use of big data enables the media not only to personalize their offer of information more precisely, but also to promote an innovation process that tends to increasingly combine journalistic know-how with technical skills unrelated to journalism. Even if the profession does not disappear, it undergoes an evolution by refocusing on some of its fundamental principles.