How does one lead newsrooms in the age of generative AI? 

Here are some interesting use cases and learnings I came across in a report published by the European Broadcasting Union (EBU), an alliance of public-service media that counts 113 members in 56 countries. You can read here how four new brands are using GenAI to better connect with their audience.

Radio Sweden and Bayerischer Rundfunk are trying to use AI to solve a genuine problem — and the motivation behind most of these experiments is that they are trying to serve their audiences more effectively.

Still, the role of humans remains incredibly important.

Radio Sweden produces and publishes 370 audio news clips every day. Each clip needs a headline, a summary in three bullet points, and an alt-text (for image description). So, Radio Sweden built a tool that creates these texts, at a high standard of accuracy, from audio transcripts. An editor then looks over them before publication.

Still, the headlines and bullet points “are very standardised and lack the more elegant style added by a skilled editor,” the report points out, adding: “It’s of paramount importance to include the newsroom from the start — to get their acceptance for using AI in the sensitive editorial workflow.”

German Public broadcaster Bayerischer Rundfunk is experimenting with personalisation, including an interactive audio news brief that focuses on a certain geolocation, which the user can either type in or which could automatically be located. 

The listener can then customise their news brief — for example, asking for news within a 50-kilometre radius, no older than 24 hours, and an alert each time a new item arrives. The AI system then creates a podcast for that listener, who gets an alert and can listen to it, like any radio programme. 

“We have very good user feedback on this product because people are really interested in regional news, and they also want to customise and personalise the news for their needs,” said Uli Köppen, who heads the AI and Automation Lab at BR.

BR has also rolled out an AI writing assistant that helps journalists be more effective, for example, by creating versions of a story for different radio programmes, its Web site, and for TV. 

“And we are debating on how to use this tool. Like, is it just an assistant? What does ‘just an assistant’ really mean? How much text do we produce with it? And do people still have to work on those texts? We’re still saying yes. 

“There are cases where we do have direct automatisation. But in most cases, people should use it for getting ideas, supporting the creative process, for versioning and then reworking the version. We usually don’t want to replace any kind of decision-making process with AI. We carefully review results generated by generative AI because of hallucinations. We can’t afford to publish any mistakes.”

How should newsroom leaders best formulate an AI strategy? 

The EBU makes a strong point:

“They need to own two things: their journalism and their audiences. They must be driven by the desire to invest in, produce, and resurface engaging and original journalism, and the desire to connect with audiences and communities. These two needs should drive any strategy that is supposed to contribute to a healthy media ecosystem, no matter which latest technology is used as a means for that.”

This journalism and audience strategy must be complemented by a tech strategy and a talent strategy:

“While investing in tech talent is important, investing in journalism talent could become even more so in a future where originality of content and human skills of delivery and connection with audiences are key success factors. News organisations increasingly bet on journalists as personal brands that convey authenticity, competence, and, well, humanness  crucially important when automation, cloned voices and avatars are the alternative,” the EBU said.

News brands must invest in reporting — and in critical thinking. 

“Journalists of all trades and fields of expertise need to sharpen their abilities to question the output AI generates. While the old reporting advice ‘follow the money’ is still valid, it will become increasingly important to ‘follow the data’ to uncover which input has shaped AI-generated output and which parts are missing.” 

If you’d like to subscribe to my bi-weekly newsletter, INMA members can do so here. 

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .