« This op-ed marks Media Off-Site's first public statement, a confidential community of media executives founded in June 2024.
Our group meets regularly, away from the spotlight and conferences, to discuss the future of our industry with complete candor. During these gatherings, we share our doubts as much as our experiences.
We will publish these insights regularly, hoping they will enrich discussions within your media teams, as well as our own.
Feel free to share this article and send us your own thoughts and feedback.
Happy reading. »
Max Leroy — President and Founder of Media Off-Site
There's always a big gap between what people say in public and what they say in more confidential settings… This is probably even more true when it comes to AI, a major strategic issue for all media organizations in the years ahead. To discuss this topic, the confidential framework provided by Media Off-Site — an exchange and mutual support space for media executives, where both their identities and discussion details remain confidential — is invaluable. It has enabled us to develop a set of six best practices, which we're sharing with you today, reflecting discussions that often diverge from publicly stated positions.
Yes, we must defend the intellectual property of media at all costs—the fruit of editorial work and ethical standards that platforms lack. Yes, we must also harness AI in our media to improve our processes, enrich our journalistic practices, and imagine new editorial formats. Yes, we must experiment, launch projects, and necessarily fail. Yes, it is necessary to explain, support, and train our teams to carry out these transformations with minimal disruption.
But it's far from sufficient!
Bet on acculturation grounded in practice, rather than formal training
In particular, many of us view training as "a necessary step for social peace", yet often find it too theoretical and disconnected from real needs, based on textbook scenarios rather than the concrete challenges newsrooms face and experimentation. "I don't believe in training. I believe in acculturation", as one of our members put it.
This prior acculturation is all the more necessary because in practice, initial good intentions are often held back. In one of the newsrooms represented, the few journalists participating in the working group on artificial intelligence thus prefer to remain discreet, fearing they will be perceived as "traitors" by their colleagues…
Create a multidisciplinary group to support AI innovation
And yet, such cross-functional working groups are far from existing in all media groups. Still, on AI as on all innovation and change management topics more broadly, entrusting the subject to a single person simply doesn't work. It goes without saying that conversely, entrusting it to no one doesn't work any better.
"Everyone's passing the buck," noted one participant, while another summarized the situation this way: "We tell ourselves 'it's a tech issue,' but at the same time it's also an editorial issue… So it's kind of everyone's issue, because everyone knows we need to deal with it, but at the same time it's nobody's issue."
In this context, establishing a clearly identified cross-functional team tasked not only with addressing ethical and deontological issues but also with driving innovation itself appears to be an essential step. However, this should not involve creating an unwieldy bureaucratic apparatus.
Be realistic and pragmatic: the best way to avoid disappointment
Even for media outlets that have already advanced significantly in their experiments, initial initiatives often reveal a major gap between initial expectations and reality, which can be a source of disappointment and, sometimes, disinvestment.
As one member noted, "everyone has the impression that in three months we'll get amazing things done […] but it's far more complex than expected". Advance work on prioritizing and assessing project feasibility is therefore necessary, while remaining aware that obstacles are just as much organizational as they are technical. "It's often the projects born from personal initiative that see the light of day fastest. A developer builds something over a weekend and shows up Monday with it, demonstrates it to everyone, and a week later it's in production".
Regardless of where the initiative comes from, one of the keys to successfully implementing an AI project or pilot lies in its alignment with a genuine "user need".
We need to ask ourselves the question of transparency, even if there's no straightforward answer
True, the technology is not yet 100% reliable and falls far short of fulfilling all its promises, but the first use cases — automatic article translation, summary generation, or analysis of reader comments — are very much already here. And with them come inevitable ethical issues, starting with the question of transparency toward audiences.
Specifically, should the use of AI tools always be mentioned? On this point, views diverge depending on the use case. While AI use for image generation or comment moderation is generally made explicit, its use as a writing assistance tool sparks far more debate: how is that fundamentally different from a spell-checker?
Some participants thus advocate for a very pragmatic approach: for them, AI is simply one assistance tool among others in the journalistic production pipeline. From this perspective, using an LLM to facilitate writing isn't fundamentally different from other writing aid tools, as long as the final result is validated by the traditional editorial chain. Other participants, meanwhile, take a stricter position on the need for transparency.
These questions, quite present today, could nonetheless become less sensitive in the future… "These are very important questions today […] But perhaps they won't be an issue in 5 years", says one participant.
Improve content discoverability through AI, rather than produce more
It's also worth noting that certain types of AI uses encounter less resistance than others. Its application to article discoverability is generally welcome, as it helps enhance the value of editorial work. The question is particularly topical for media outlets that produce a large volume of content.
As a senior figure at a major media outlet acknowledged, "we produce 130 pieces of content a day. It's absolutely depressing how much of it goes unseen." AI could thus make it possible to better expose content, by personalizing the experience on websites and apps…
Enhance content value by "repackaging" it with AI
Along these lines, some participants also anticipate the emergence of new roles, such as that of "augmented editor" capable of mastering AI tools to adapt content across different formats and channels: "We may soon see a new profession emerge, coming after journalism and content creation in general: content adaptation".
These "augmented editors" would be capable of mastering prompt engineering, refining summaries and ensuring alignment with editorial guidelines to adapt content for multiple distribution channels — across different social networks, but also in multiple languages, through adapted subtitling, or even with automated dubbing.
"That's what all YouTube content creators and others do with tools that let them re-cut content for vertical or horizontal video, for instance", noted one participant. But the practice is still far from being established within "traditional" media outlets.
N.B.: Marine Doux and Baptiste Thevelein, co-founder and co-founder of Médianes, also co-founded Media Off-Site. This article is also published on LinkedIn, and on The Audiencers.
