Editor’s note: This article was created in partnership with smartocto, a smart editorial analytics system.

AI has been top of mind for most news leaders over the past few years. Research shows that most news organisations are already using AI tools for tasks like back-end automation and optimising content creation.

Erik van Heeswijk, CEO of smartocto, notes five key categories of AI use in media:

automatic content production (such as text or image generation);
story assistance and rewriting (e.g. offering SEO insights);
content classification, such as recognising user needs;
decision making (such as offering analytics and insights like when to publish a story);
distribution – personalising how the content gets to the audience.

Not all publishers embrace all categories immediately, partly because of concerns around issues like copyright and hallucinations. However, these objections can be alleviated by using AI tools designed specifically with news publishers in mind, giving media organisations full control over their data.

We looked at three practical use cases for publishers that underline the combination of editorial data and the added value of AI.

Automated user needs analysis

As with other products, different users have different needs when they are consuming news. Someone is coming to your newspaper’s website to learn what’s going on in the world, another person wants something light and entertaining, yet another reader needs actionable advice.

Close to a decade ago, the BBC World Service Audience team pioneered the user needs concept that has since become a widespread trend in the publishing industry. Dmitry Shishkin, its most prominent advocate, together with smartocto recently came out with an updated version – User Needs Model 2.0 that identifies eight different reasons why people consume news.

*A visual showcasing User Needs Model 2.0 and one of its developers Dmitry Shishkin (image: courtesy of smartocto)*

This approach helps publishers increase audience engagement and loyalty, as well as attract and retain paid subscribers. One important point is not focusing too much on producing straight-up news stories and making sure to keep other user needs in perspective.

Many newsrooms “make too many ‘update me’ news stories, and they have to consider these content strategy models to make different content and to align it more with the user needs”, Van Heeswijk says.

To move beyond the theoretical concept to its practical use in the newsroom, publishers need a process that takes into account their audience data and automatises the application of different user needs in the day-to-day reporting process. That’s where an AI tool is indispensable.

Smartocto offers a free public beta tool that allows everyone to analyse their article and see which user needs it addresses. Clients have access to a more sophisticated system that allows to streamline the employment of the user needs model in their newsroom. Contact smartocto’s team to get a free consultation.

Choosing better headlines

Major publishers like The New York Times famously A/B test multiple versions of a headline for each article to see which one will do better. Smaller news outlets don’t have the luxury to do this at scale but still put a lot of care into choosing a good headline that would do justice to the story and attract audience attention.

This can be a difficult and time-consuming process. Thus, headline generation is one of the most obvious applications of AI. After all, today’s AI systems are not capable at generating high-quality content from scratch but excel at language tasks that involve repackaging existing text.

Forbes.CZ developed an internal AI tool that “can offer its own suggestions based on successful headlines from past articles on the same or similar topics, and any specific internal editorial guidelines”. Top business outlets like the AP and Bloomberg have been experimenting with using AI for simple financial stories for years now.

For smaller newsrooms, ChatGPT has emerged as a good alternative. The experience of Dutch broadcaster Omroep Brabant shows that “if the editorial team lacks the time or energy for brainstorming alternatives, ChatGPT can be a valuable collaborator with just a few clicks”.

However, ChatGPT has some serious limitations, including its tendency to generate clickbait-oriented headlines with too many emotional words that don’t necessarily align with the publisher’s style. Tweaking the prompt and giving feedback, while time-consuming, might help. Another tip is making use of custom instructions available in ChatGPT’s paid plan.

Smartocto offers headline generation among its services. Unlike generic tools like ChatGPT, the system is trained on a specific publisher’s data and allows adjusting the headlines to different goals like conversions, loyalty or engagement. Headlines it generates will be automatically written in the style of your publication. Request a demo to learn more.

Creating a virtual assistant that helps make better decisions

One way to think about AI tools is that you get access to a personal assistant. Just like a human assistant, it wouldn’t do your job for you but can help make you focus on the most important tasks and make them more impactful.

As DeepMind co-founder Mustafa Suleyman told CNBC, over the next five years everyone will be able to have their “own chief of staff” that will be able to “reason over your day, help you prioritise your time, help you invent, be much more creative”. There are already use cases today with news leaders using AI tools to help them build their personal schedule or prepare for important meetings.

The key is to experiment with AI tools to see how they can offload routine tasks – and to understand their strengths and limitations. As Ethan Mollick wrote comparing AI with an intern, “just like any new worker, you are going to have to learn its strengths and weaknesses; you are going to have to learn to train and work with it; and you are going to have to get a sense of where it is useful and where it is just annoying”.

*Example of the roles a virtual assistant in the newsroom can play (image: courtesy of smartocto)*

“We think of AI as basically a virtual assistant to every role in a newsroom” from editors to business analysts, smartocto CEO Erik van Heeswijk says. For example, a chief editor can get advice on which types of stories produce more of and who among reporters needs more training, while an analyst can get ideas for getting more advertising value. Read more or request a demo.

Three ways to combine data and AI in your news organisation

The Fix Studio

Advice on using AI to improve publisher operations

Author

The Fix Studio

‌

‌‌‌

‌‌‌