During this week’s INMA Media Innovation Week in Helsinki, attendees got a quick dive into four topics key to the news industry via seven, seven-minute presentations. 

Here are seven case studies in AI presented at the conference: 

Tamedia trains teams on AI

Tamedia in Switzerland offers a prime example of how companies can train teams for AI proficiency. Nadia Kohler, the AI Lab lead at Tamedia, formed the in-house lab just a year ago. Since then, her team has successfully built and integrated an AI tool for nearly 600 journalists across Switzerland. 

This process, using what Kohler called the “AI pyramid,” is composed of three key stages:

• Exchange at the base.

• Testing.

• Rollout at the top. 

“You cannot roll out before having exchange and testing,” she emphasised, underlining the importance of collaboration and experimentation before full implementation. 

To engage employees, the team experimented with various outreach methods and found that a simple old-school flier proved most effective. More than 640 employees scanned the QR code on the flier to enroll in the AI course.  

Kohler described AI as “a people’s business,” and part of her strategy was to integrate employees into the development process, allowing them to test the tools and provide feedback. “Let those affected experiment,” she advised. 

These early adopters became internal advocates, encouraging their colleagues to explore AI as well.

Effective communication was crucial to the lab’s success. The team ensured all AI-related communication was accessible by providing it in three different languages and using multiple channels. They also marked AI communications with a mascot, making the information more engaging and recognisable. 

Understanding that every employee learns differently, the lab provided both in-person training and an online course. Additionally, AI hackathons allowed employees to experiment with the tools in a casual, supportive environment — complete with snacks and drinks to create a welcoming atmosphere.

Kohler reiterated the importance of frequent feedback: “Ask your employees how they feel regularly and take that feedback seriously,” she advised. 

In just 12 months, the AI Lab introduced 20 AI-based tools to 600 employees, reduced AI skepticism by 5%, and increased the number of AI enthusiasts within the company by 15%. 

Russmedia’s AI Future Team focuses on education

In Austria, Russmedia has introduced an AI Future Team to support the widespread adaptation of AI. Chief Technology Officer Dominic Depaoli explained how Russmedia strategically tackles AI integration through three core initiatives: surveys, tools, and testing. 

The team conducted surveys to understand employee expectations, hopes, and concerns around AI, ensuring these insights guided their approach. The tools initiative involved identifying essential AI tools, testing them, and developing custom in-house solutions, including GPT models, tailored to Russmedia’s unique needs. This enabled them to streamline workflows and enhance content production.

Depaoli stressed the importance of education in effectively integrating AI into the workplace. In 2024, the AI Future Team organised over 70 workshops and created prompt catalogs to help employees familiarise themselves with AI tools. 

Evaluation shows most employees now use AI daily. He likens AI to a “super car,” urging caution: “Give employees go-karts first, let them practice and make mistakes.”

At Russmedia, the belief is AI can enrich nearly every story. With their strategic focus on employee input, tool development, and comprehensive training, the company has created an environment where AI enhances creativity and efficiency without overshadowing the human element.

Kölner Stadt-Anzeiger increases engagement with AI recommendation service

Germany’s Kölner Stadt-Anzeiger (KStA) has successfully transitioned to an AI-driven, fully personalised front-page experience. Robert Zilz, head of data, outlined the roadmap for this development, which began with small AI integrations (such as cloud services and speech recognition) in 2017 before advancing to recommender systems and full personalisation. 

While AI itself isn’t new, the possibilities it offers continue to expand, Zilz said. KStA approaches AI as the intersection of products and applications, processes, and organisational culture, with AI sitting at the centre. 

A major breakthrough came with the introduction of its AI-powered recommendation service, which increased click-through rates (CTR) by 80% by delivering tailored content to users. This also boosted the number of fully read articles by 13%. Currently, 80% of KStA’s front page is curated by AI, with the remaining 20% managed by the editorial team.  

The recommender service, designed specifically for KStA, suggests both articles and topics, significantly improving user engagement.

Looking ahead, the goal is to fully automate the front page, eliminating the remaining 20% of editorial curation. He advised companies not to aim too high too soon: “Don’t reach for the stars first.” 

He stressed the importance of understanding data and investing in AI-ready infrastructure before diving deep into AI technologies. Furthermore, Zilz encourages using “plain AI” where possible, cautioning against overreliance on generative AI alone. His final piece of advice for anyone using or experimenting with AI is: “Don’t overestimate or underestimate the power of AI.” 

Financial Times retains subscribers with automated save offers

One key strategy at FT is delivering the right offer to the right user at the right time. With that, they want to get new valuable subscribers as well as keep the older subscribers, explained Maria Sakki, data lead analyst at the Financial Times. 

In 2022, when a subscriber was about to cancel their subscription, the save offer recommendations were based on the customer’s current price. Now, the save offers are rule-based and are based on factors such as price and user profile. In the future, the offers will be dynamic, AI-driven recommendations.  

FT estimates that a 1% increase in retention is approximately equal in value to a 15% increase in acquisition.

What they have learned so far is that the increasing use of large language models (LLMs) applications is diversifying the user base, which strengthens expertise across teams. Furthermore, more teams are using LLMs for little tasks, such as summarising articles.

JP/Politiken Media Group uses internal platform for GenAI tools

JP/Politiken Media Group’s internal GenAI product Magna offers a safe chat environment within the group, keeping data for its employees and not providing it to third parties. 

Tore Rich, product manager/GenAI at the news group that works in Denmark, Norway, Sweden, and Germany, explained the product allows journalists to upload their articles and get feedback. Magna uses generative AI for routine tasks and to enhance news coverage through text, images, and audio. 

Moreover, Magna offers an archive of articles, where journalists can search for previously published articles. It will also provide headline suggestions and proofreading for the news team, as well as eight tools designed to assist the editorial staff — a bullet point summary, proofreading, and spellcheck. It also can shorten the article by removing the less important parts.

One of the important key learnings when using AI, Rich said, is that GenAI must be based on external facts when used in news. 

Ringier AI weather assistant optimises automated content

AI Boost, made up of four chatbots running simultaneously, is the brainchild of the Swiss-based Ringier’s cross-functional development team. The chatbots are text-to-voice, AI summary, conversational experience, and data-to-content.

 “Utilisation is the hardest part — not building it,” said Bernd Volf, chief technology officer/chief product officer at Ringier. “It’s really getting people to use it.”

To do so, the team narrowed its focus to three priorities: relevance for users, high quality, and uniqueness. 

Weather fit all three.

“We can cover the current weather, future weather, but also historic information around it,” Volf said. “We give a bigger perspective regarding climate impact plus tangible information like tips for trips.”

Within two weeks of developing the app, daily readers using it consistently grew 7%. Next, the team plans to roll it out to the company’s other 18 titles, while personalising and monetising it.

RCS Media AI-powered reader assistance serves subscribers, archives

The starting point for RCS Media’s AI Reader’s Assistant was this question: How can the product support the growth of 650,000 subscribers? 

Daniela Buoli, head of product development at the Italian news company, explained that the company decided to expand its portfolio offering an exclusive AI app package for subscribers.

“We tried to develop an assistant that truly matters to our readers,” Buoli explained. “AI has become core to our product strategy. People pay for concrete value to solve a real problem.”

The app fosters a direct relationship with subscribers, offering  answers to their search questions sourced from 30,000+ archived articles, Buoli said. 

“The audience no longer needs to go browse the Web to find a valuable answer.”

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