Most of the early AI work in newsrooms sat behind the story: transcription, translation, a faster first draft. That work matters. It is not, on its own, why a reader pays.
WAN-IFRA’s account of Les Echos and Le Parisien, from the Paris AI Forum, is useful because both newsrooms are pointed at the subscriber, not only at the desk. Violaine Degas, chief digital officer at Les Echos, put the aim plainly: engage subscribers, because that is what retention depends on.
Les Echos built a habit, not a gadget. “Playlist de l’actu” turns the top stories into audio, updated twice a day, and keeps it for premium subscribers. “18-20” is a short evening selection, a dozen or so pieces rather than the whole river of the site. The company has also been testing generative translation of that selection for international readers, and it is building an economic-intelligence engine so a professional subscriber can query more than three decades of archive before a meeting. Le Parisien used AI to match articles to its video catalogue and insert a relevant clip, and it is building a response engine that answers from the archive with sources and links back to the pieces.
None of this replaces the article. It changes the job the article does. A commuter gets audio. An evening reader gets a cut, not a homepage. A professional gets an answer with the reporting still attached. The subscription starts to feel like a service you return to, which is a different product from a pile of URLs.
The design choice underneath is the one worth copying. These features are narrow, they sit on journalism the newsroom already made, and a human organisation still decides what “the news” is. The model is doing format, retrieval, and matching. It is not being asked to invent the story.
That is the conversion path we think is real: use AI to make the existing journalism easier to enter, easier to finish, and easier to come back to. Efficiency in the newsroom pays for the time. The audience-facing layer is what the reader notices.