AI isn't the problem...

There's something wrong with the debate about AI in journalism right now, and it isn't because the criticism is groundless, because it isn't, but because it misses the mark.

When a culture minister wants to strip AI media of every penny of support, and when cases of fabricated quotes and plagiarised stories fill the trade press, I don't think the technology is the problem, but rather the way we use it.

Take the recent case from Berlingske, where a journalist was sent home because an AI tool had fabricated quotes, passages and entire people in an article about the future of drinking water, and where the experts being quoted either couldn't be found or had never spoken to the paper. Berlingske's own editor-in-chief put it very precisely when he stated that AI must not write their articles, and that a human must process everything the tool delivers, but both rules were broken here.

For me, this is therefore not a story about AI being dangerous in itself, but rather a story about a newsroom that didn't have a handle on its own AI strategy. And that's exactly where I'd like to step in, because I fundamentally disagree that AI as such should be journalism's enemy.

The lazy copy machine that scissors out other people's stories and lets the machine write away without a human at the helm clearly deserves all the criticism it gets. But that's a caricature of what the technology can actually do, because used thoughtfully, AI gives us possibilities in editorial environments that we've never had before. I know this because I build them myself.

Alongside my job, I run the hobby project IssueAI, where I build real tools for real editorial problems. And when I look through my own work from the past few months, I can see that it touches on three areas that quite precisely show what I think AI should be used for.

The first is about organising information and making sense of complex material. Just think of the big summer events like the People's Meeting on Bornholm, Roskilde Festival and Smukfest in Skanderborg, where thousands of names of performers, debaters and celebrities flow through programmes, attendee lists and press coverage, and where you as a newsroom need to be able to figure out who is really worth pursuing.

That's why I've built a mini-app for each event that makes it possible for those covering it to find the information. After all, no one sits and does it by hand when an entire festival programme has to be searched before deadline. And even though it's AI that makes all that work possible, it's still me who decides every single rule for how the data should be handled and what you can do in the app.

The second is about planning and executing editorially, because one of the problems I know best from my daily life as a digital editor is the pressure to have an overview of everything from daily lists from Ritzau and emails that need prioritising, to our own events and the stories spotted during the day that need to be held onto.

That's why I'm building an editorial nerve centre with a single input window where, from both mobile and desktop, I can dump raw text that then automatically gets sorted and placed in the right spot in the system. The point is that it isn't AI making the editorial decisions, but AI creating the overview so that I can make them.

The third is about understanding and reaching digital users, and here I've built a web analytics app that pulls our GA4 data into the editorial nerve centre, which every morning serves the newsroom a finished analysis. That means we avoid the manual CSV export I used to do and rarely had time for. On top of that, I've made in-depth traffic analyses of our big bets, for example the coverage of the Marius case, to understand what really holds readers and why. For me, this isn't about guessing what will become a click, but about actually knowing what works.

The common thread in all three tracks is the same thing Berlingske's editor-in-chief called for, namely a human at the helm. Because while AI organises, sorts and analyses, it's me who assesses, decides and stands accountable. For me, that division of labour is the whole difference between the case that sent a journalist home and the work I do every week together with my newsroom of talented colleagues.

That's why I'm also in no doubt that AI in journalism is no longer something we can opt in or out of, but has become a necessity. Those who flatly reject the technology, or conversely use it without the thoughtfulness I'm calling for here, will fall behind both in terms of their own possibilities and their own value in a modern editorial environment. And not because the machine takes over, but because those of us who learn to use it wisely will be able to do more, faster and with a better overview than we could before.

The criticism should therefore not be about whether we should use AI, but about how. And that's a debate I'd much rather have.