Established in 1889, DN is one of Norway’s most respected business newspapers, providing comprehensive coverage of financial markets, corporate news, and economic trends. Hanne Brønmo Innerdal, DN Media’s AI Lab Lead, and Magnus Aabech, DN Media’s Development Editor, outline how the publisher is currently using AI. This feature was originally published on the Bright Sites newsletter and is re-published with kind permission.
DN Media, or Dagens Næringsliv Media, is a prominent media company in Norway, known primarily for its flagship publication Dagens Næringsliv (DN), which translates to “Today’s Business.” The publication is often compared to international counterparts such as the Financial Times or The Wall Street Journal, serving a readership that includes senior professionals, business leaders, and policymakers.
Part of the Schibsted Group since 1991, DN Media has expanded its operations to include a range of digital products and services including online news portals, subscription services, and specialised financial tools aimed at providing up-to-date information and analysis. In this interview, DN Media’s Hanne Brønmo Innerdal and Magnus Aabech outline how the publisher is harnessing the capabilities of AI.
Please introduce yourselves and tell us a bit about your roles…
Hanne Brønmo Innerdal: I’m the AI Lab Lead at DN Media Group. In January last year, we established this Lab responsible for creating editorial products and services in the AI space, but we also work with other parts of our media organisation, like the commercial and technical teams. We’re a small team but collaborate with other product teams as well.
Magnus Aabech: I’m the Development Editor at DN, the leading paper in our organisation, which covers business, society, and politics in Norway. I lead a small team of data journalists and prioritise development tasks for DN. AI is increasingly a big part of our focus, both in journalistic tools and products for the end user.
What’s your discovery process for AI labs? How do you decide what to build?
Hanne: We use the AI team in the newsroom actively. Since we’re a small lab, we asked them to gather use cases from the entire organisation—they collected around 40 and prioritised the top ten. We then discussed these with them, ensuring the list was representative of the newsroom’s needs.
We also consider technical and ethical aspects, as well as whether some suggestions might be more suitable for long-term projects or something we should buy instead of build. Top management expects us to deliver at least three services within a year, so we also have to look at what’s realistic to deliver within that time frame. The closer you can be to the newsroom, the better, but it’s important to have open dialogue as well.
What are some specific ways that your newsroom is currently using AI?
Magnus: The first thing we created was transcribing audio to text. That was our first experience with generative AI, although we’ve been working with AI for several years, recommending articles, robot journalism, and using natural language processing for automated tagging, etc. Then we started working on a summariser.
Hanne: We saw that Schibsted started experimenting with a summariser—an AI-generated short summary of the article at the top—which had a positive effect on engagement and reading times. We were curious to see if we did the same, how that would affect people accessing our journalism. As a business newspaper, our articles often cover complex subjects and are long, so finishing time isn’t always high.
The summariser creates a version of the article in a box, which a human has to approve. You can have three or five bullet points, or ask it to summarise simpler or focus on different aspects of the article. This was the second product we built about a year ago. We’ve taken that further by working on creating an AI module in our CMS, adding title suggestions, and experimenting with fact-checking or using AI as an editorial assistant for reporters. We’re also exploring other features where AI can help reporters improve an article.
I’m really interested in this summarisation helping engagement because it feels like reading the summary would decrease engagement, rather than increase engagement. Can you tell us more?
Hanne: Our initial analysis showed that while the number of people reading the whole article went down slightly, engagement in terms of reading more articles went up—there were more articles read per visit. This is crucial because we want people to read more when they visit. If they read four stories instead of one, that’s high value and helps us pay for our journalism.
We haven’t implemented this for every type of article. For instance, long feature stories or portraits are harder to summarise. But for many news articles, it’s a valuable feature, at least according to the readers.
Magnus: Especially with a younger audience, there was much higher usage of the expand button. Younger readers, a group we have trouble accessing, used it more. It’s one of the more important developments for that audience.
How else are you using AI in investigative journalism?
Magnus: AI helps us process large volumes of documents. It would take a journalist several hours to manually check everything, but AI can detect the relevant information faster. Journalists still manually go through the findings, but AI significantly speeds up the process. This was our first real example of using AI as a research tool in investigative journalism, something we’ll do much more of in the coming months. All journalists in the DN newsroom have been, or will be, given ChatGPT training, and we’ll build some tools for them directly, possibly a sandbox for experimentation.
What challenges do you face when developing AI products?
Magnus: One challenge is the need for human involvement in AI processes, especially with generative AI. We need to control it to avoid issues like hallucinations, which makes development more complex. It’s not straightforward; we have systems in place to check for hallucinations and ensure accuracy.
What’s the response been from journalists to generative AI?
Hanne: We’ve seen the whole spectrum—some are super engaged, spending hours teaching themselves AI and using it daily, while others are sceptical or unsure where to start. Most people fall somewhere in between.
Magnus: Getting people to start exploring AI is often a challenge because they’re so busy and sceptical. But overall, it’s not too bad. Eight years ago, when we started using AI in journalism, it was more difficult to get people to understand the potential. Journalists are impatient; they want tools to work perfectly right away.
A year and a half since ChatGPT’s launch, people understand more about what the technology can do. They’re eager for tools to help them in their jobs but are more sceptical about products for the end user. Overall, the response has been much more positive.
Hanne: There’s also been a shift in thinking about building versus buying AI tools. Sometimes we realise it will take us a year or two to develop something, so we collaborate with others or buy off-the-shelf solutions. These are important internal discussions.
How do you decide what to build and what to buy?
Magnus: It’s important to try building some things yourself to understand the technology. When we buy something, it often requires adaptation—it’s not just plug and play. But we do buy some stuff, depending on the tool or service.
What advice would you give to newsrooms just starting to explore the use of generative AI?
Magnus: Explore the possibilities and understand your challenges by playing around with AI. Build a custom GPT, see if it can help in your newsroom, but be aware of the challenges and pitfalls. If you want a really good tool, you might need some additional code development.
Hanne: We have a group of AI ambassadors in DN—our go-to people for discussing ideas and testing stuff. For me, as someone in the lab not directly in the newsroom, it’s important to have close connections to ensure we’re developing things that reporters actually want and find useful. Close dialogue with top management and regular reporters is crucial.
Republished courtesy of Bright Sites, the creator of the FLOW digital publishing platform which incorporates a data-driven approach combined with machine learning, AI, e-commerce, subscriptions and translation. Bright Sites developed FLOW, a software-as-a-service digital publishing CMS, that provides multi-location newsroom workflows, multi-lingual content creation and AI to personalise the user experience used by a range of global and local publishers. Clients include Irish Independent, The Independent, London Evening Standard.
