‘AI touches every layer of our transformation – from strategy to tooling, culture, audience – but it also builds upon everything we’ve already done. In my opinion, it’s not a rupture – it’s the continuation.’
Based in Munich, Süddeutsche Zeitung (SZ) is one of Germany’s largest news publishers, with more than 500 journalists and global correspondents, and more than 300,000 digital subscribers.
SZ started their AI newsroom integration about a year and a half ago, by establishing an AI Board of experts to coordinate all AI initiatives. They also set up an AI Forum with experts from various departments: IT, HR, legal, and the visual desk. These, along with their product teams, data scientists and a growing network of AI-savvy editorial experts, forms the structure that manages SZ’s AI transformation.
As SZ’s Managing Editor and Senior Editor for AI, Fabian Heckenberger leads the newsroom’s strategy team, with a growing emphasis on the journalistic applications of Generative AI (GenAI).
Flipping the standard transformation script
“GenAI didn’t start the transformation in media, but it accelerates, and probably changes, everything,” he asserted at WAN-IFRA’s recent Congress in Krakow. In a session on Rethinking Future Newsrooms, Heckenberger illustrated how the agile publisher approached AI integration by looking back at decades of digital newsroom transformation – and realised that their lived experience did not align with the standard linear model.
“You know the story: from left to right, one phase leads to the next…There’s a roadmap, a starting point, and after a while, voila, you’ve transformed,” quips Heckenberger.
So SZ flipped the script, by adding a change model to every layer of the four stages of transformation. “We think of it as a layered model; a change foundation model, that got us so far to over 300,000 digital subscribers.”
The four stages of transformation
1) Strategy. Define your North Star: “We went from serving two gods – subscribers and advertising – to putting subscribers first.”
2) Tooling, Tool-driven change. Updated CMS architecture: “You really embrace the force of tool-driven change when you implement the right tool, at the right moment, for the right people.”
3) Culture. Start small: From theory, to best practice, by asking the right questions – “What do people in this newsroom really need? And what do subscribers really want from us?” – and building a “change muscle”.
4) Audiences. User needs model: “Rethinking what content our users, our audience – not a single audience, all the different audiences we have right now – really want and need from us; and going back from creating more content, to creating more value – and to being more essential for our readers and our subscribers.”
Transformation reimagined, with AI at every level
For most, integrating AI implies simply adding another layer at the top of these stages. But, stresses Heckenberger: “GenAI doesn’t just sit on top, in my opinion – it cuts through every layer of this transformation. It reshapes strategy, tools, culture, audience relations, everything. Every decision.”
Strategy redefined
The North star remains affixed, but the strategy now also includes tools to improve workflows, strengthen the journalism, and reach new audiences – while supporting subscribers.
Tool-driven change, from the bottom up
SZ uses the Langdock CMS platform to allow teams to safely experiment with different LLMs and share them across the newsroom.
“This way we see what journalists are actually really using, what they actually need to build for themselves, and then decide what to scale and embed.” And what to quit, adds Heckenberger:“If an AI project or initiative doesn’t fit one of these goals, it doesn’t happen.”
Building a culture of trust: Expert in the loop
Transparency is key here: ZS publishes their AI guidance for readers, subscribers and internal staff every six months. “This is very important because this helps to build trust, to tell people how we in the newsroom, as journalists, use GenAI and how we don’t use GenAI,” adds Heckenberger.
To establish trust as an essential baseline, SZ redeveloped the concept of ‘human in the loop’ to ‘expert in the loop’ – by implementing the expert from the start, through the entire process.’
Heckenberger illustrated this with how they implemented article summaries by working closely with journalists, first to “define what an Süddeutsche Zeitung style summary looks like; the tone, the structure, the length, the context, the format,” he explained.
“We ran a prompt engineering side by side with the authors for hours using their real articles. And this was time well spent: the editors now trust the tool and the feature; the subscribers benefit from better GenAI summaries – and we establish, step by step, the trust within the cultural layer.”
Audience focus – and shifting perceptions
Building dedicated, audience-focused tools has benefited SZ’s transformation both internally and externally, notes Heckenberger. He shared how an AI-powered assistant that SZ built during Germany’s recent national elections, to help users navigate party platforms, compare positions and explore policies, fed back invaluable audience insights.
This data flowed from analytics into editorial planning. “It was even used to create news stories and again helped to develop trust within the newsroom for this new technology,” adds Heckenberger.
With results like these, it’s clear why he views AI as a transformative force. “AI touches every layer of our transformation strategy to tooling, culture, audience, but it also builds upon everything we’ve already done. In my opinion, it’s not a rupture – if you want to be optimistic, it’s the continuation.
For newsrooms still hesitant about their approach to, and integration of, AI, Heckenberger posed three pertinent considerations.
The digital transformation and the GenAI transformation overlap: is this a risk or an opportunity?
Is the change muscle within your organisation strong enough to deal with these transformations at once?
And: if you think about the identity of the user, the identity of your media organisation as the sum of all small workflows, of all small roles and these workflows and roles are now changing, what does implementing more and more AI mean for the identity of your journalistic approach and for your newsroom organisation?
