Ippen Digital is building what it sees as an AI-native newsroom, using agentic AI not just to automate tasks but to redefine how journalism is created, distributed, and scaled.

At the INMA World Congress of News Media study tour stop in Berlin, the German publisher detailed how AI agents are embedded across its newsroom — from data ingestion to content generation to distribution to quality assurance — and increasingly across the organisation itself.

The visit was part of World Congress week, which kicked off on Monday with three study tours. Coming from around the world, 94 media professionals are visiting 17 German media companies during the study tours.

Operating a network of more than 70 journalism Web sites reaching 27 million people monthly, Ippen Digital has leaned into AI as a necessity driven by scale.

Building the AI-native newsroom

At the core of Ippen Digital’s strategy is the idea that AI is not a tool layered onto journalism but a foundational system shaping how the newsroom operates.

Markus Franz, CEO of Ippen.AI — Ippen Digital’s in-house AI unit that builds and integrates AI tools across editorial and product workflows — described an emerging model in which AI functions as an operating layer across the organisation.

Central to this is the use of agent-based systems that analyse content, generate outputs, and refine results in a continuous loop. Rather than static publishing workflows, Ippen is moving toward a system in which content is constantly evaluated and improved based on real-time signals.

Franz also highlighted the growing importance of how the company is employing “skills” — lightweight, rules-based agents that allow teams to build new editorial or product capabilities quickly. This approach reduces reliance on traditional development cycles and enables faster experimentation across the newsroom.

The shift, he suggested, is not just technological but organisational, requiring publishers to rethink workflows, roles, and how decisions are made.

From production to distribution

Within this AI framework, Ippen Digital is also confronting a fundamental shift in the content value chain. A central theme of the session was that production is no longer the bottleneck — distribution is.

Using election coverage as an example, the company generated about 5,000 hyperlocal articles in a single week, each tailored to individual communities using structured data.

“Mass production is not the problem anymore,” said Markus Knall, editor-in-chief at Ippen. “The problem is the distribution.”

Rather than overwhelming audiences, Ippen distributes content across its network so each reader encounters only the article relevant to them, Knall said: “A single reader would never see those articles. He would see one single article that is important for him.”

This marks a shift from volume to precision — where success is defined by how effectively content reaches the right audience.

Hyperlocal content at scale

AI enables Ippen Digital to push into levels of localisation that would not be possible manually.

“My theory is that we produce very highly individualised content for local interests,” Knall said. “The smaller the community … the better it is for us.”

By focusing on small communities — sometimes just a few hundred people — the company uses structured data and automation to create highly relevant coverage at scale.

“I would be happy if we could produce [an] article for maybe a single person,” Knall added.

This hyperlocal strategy is positioned as a key competitive advantage against global platforms that lack access to granular, proprietary data.

A new economic model for journalism

AI is also reshaping the economics of content.

While individual AI-generated articles may generate modest traffic, their low production cost allows Ippen to scale output profitably across thousands of topics and locations.

AI is changing the “I” in ROI for newsrooms because “the investment is getting smaller,” he said, referring to the cost of human-produced content vs. AI-produced content. This model shifts the focus from maximising performance per article to optimising return across large volumes of content.

Quality and trust in an automated system

Despite the scale of automation, Ippen executives emphasised that editorial standards remain central: “This is an election, and there can’t be any mistake,” Knall said, referring to the safeguards built into its AI workflows for sensitive coverage.

Journalistic guidelines are embedded directly into AI systems, and the company invests heavily in data validation to ensure accuracy.

Transparency is also part of the model. AI-generated articles are labelled, and readers can verify the underlying data.

“If you don’t trust this article because it’s AI written, that’s fine for me,” Knall said. “Then click on the map and control all the data.”

Executives suggested concerns about AI authorship are more pronounced inside newsrooms than among audiences, where trust is more closely tied to accuracy and usefulness.

What news executives should take away

The Ippen Digital visit offered a clear message for news leaders navigating AI transformation:

• AI must be embedded across workflows, not treated as a standalone tool.

• Distribution — not production — is becoming the primary constraint in digital publishing.

• Hyperlocal, data-driven content offers defensibility against global platforms.

• AI changes newsroom economics by lowering costs and enabling scale.

• Becoming AI-native requires organisational as well as technological change.

Ippen Digital’s approach underscores that the future of journalism will not be defined by how much content publishers can produce but by how intelligently they can use AI to connect that content with audiences.

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