Martin Schori spent 13 years at Aftonbladet, Scandinavia’s largest newsroom, rising from a news editor to director of innovation and AI and deputy editor-in-chief. For two years, he ran the paper’s AI hub, a team of journalists, developers, and a designer. His verdict on the media industry’s early approach to AI — laid out in a new book, AI in the Newsroom — is blunt: newsrooms got it wrong. They chased efficiency, building tools for reporters who never asked for them.
Brand publishers fielding a similar mandate from leadership — use AI, cut costs, and move faster — now face the decisions Aftonbladet confronted several years ago, with a fraction of the budget. Schori’s advice: start where the value is, win over the people who run day-to-day operations, and build tools that are too useful to ignore.
This interview is part of a Brand Media Review series featuring the editors, CMOs, and content directors at the forefront of brand publishing.
You’ve said the media industry got AI wrong early, treating it as an efficiency play. Where’s the real value?
We were very focused on efficiency, but we should have started with other parts of the company. Most AI strategies in news media are newsroom strategies, and if you’re hunting for efficiency, that’s a natural place to start. But the business side — churn prevention, marketing, and new interfaces like chatbots for news — is where I see the most value now. And agents, obviously.
Isn’t that an uncomfortable answer for editors whose bosses are pushing them to integrate AI into their work?
Efficiency makes sense if you’re the owner of a company, or a CEO. For a reporter, being more efficient is not something you dream of. You might not see the point of it at all. So, there have to be other incentives.
You built a “buffet of AI tools” for the newsroom, and most of them failed. They sat outside the CMS, and the friction put people off. So, what does getting those incentives right look like?
Start where there’s a problem, where there’s value you can add. We started with what was possible to build, rather than where the problems were in the newsroom’s processes. The technology also wasn’t very good yet. One lesson we learned: we worked a lot with the reporters and journalists, but we didn’t engage the middle managers, the ones leading the daily operations. If they’re not engaged, nothing happens.
Our readers are building editorial teams and newsrooms inside companies, rather than in legacy newsrooms. Where should smaller teams start with AI?
It depends on what you’re trying to achieve. Take a newsroom. If you want to give your target audience a smart way to get the latest news in a way that’s personalized and well-presented, you should use AI and agents, because AI is going to do a better job than humans at selecting and presenting. But if you’re a small newsroom, or a company that wants to connect deeply with your audience, with a personal tone of voice, maybe you shouldn’t have any AI at all, except in the background. There are two extremes: the really smart, fast, personalized one, and the one with humans at the center. Most newsrooms end up in the middle. You should choose one of those two extremes and become really good at it, instead of ending up in the middle.
You’ve said the workflows newsrooms spent time and money building can now be bought cheaply, and that smart editorial teams might do better to wait. How should a smaller player decide between building, buying, and waiting?
Experimenting is in Aftonbladet’s DNA, so we jumped in. We invested a lot of money, time, and resources in new tools and products, and we learned a lot. But everybody can’t do that. And a lot of what we spent months developing, you could buy off the shelf today, cheaply.
On the other hand, there will come a time when AI isn’t a competence anymore, because it’s so integrated into daily life and work — so, don’t disregard it entirely. In the beginning, there was this notion that you had to jump on board now, transform everything, or it would be too late in six months. That didn’t happen. Learn what’s going on and find use cases for your organization. But not everybody has to be the innovative newsroom spending a lot of money to build tools of their own.
Some of your findings surprised me. You’ve written that AI summaries made readers stay on the page longer, for one. What else surprised you along the way?
Quality is everything. If you ask your audience whether you should get engaged with AI, they’ll probably say no. If you ask your journalist colleagues whether you should work a lot with AI, they might also say no, because they’re skeptical of the technology. But if you give them a transcription tool that saves hours of boring work, they’ll use it. And if you provide your audience with a chatbot based on verified journalism, they’ll start using it.
We launched Hej Aftonbladet [a chatbot designed to provide answers to readers’ questions, trained on the paper’s own journalism], and it answered 7 million questions during my time there. I don’t think we got a single email about it, and we get hundreds of emails every day. If you spell something wrong, you get a hundred emails asking whether you have to be an idiot to work at Aftonbladet. With the chatbot, nothing. People just used it. The AI projects that succeeded were the ones where we didn’t get any feedback.
You mentioned transcription. Those boring jobs — transcribing and summarizing — are often how young journalists learn. If you automate the tasks that early-career writers train on, how does the next generation build judgment?
I’m not sure I have an answer, but it’s a question we need to talk about a lot more. The efficiency gains are going to come from commodity journalism. Owners aren’t going to tolerate 25 or 30 people manually rewriting other newsrooms’ stories, or the latest interest rate news. So the tasks we call simple, which were maybe never that simple in the first place, are going away, and we have to find other ways to train people. The whole craft is about to change.
This interview has been edited for length and clarity.
