Some local media organisations have been benefiting from AI for years. Peter Houston explores what it looks like in practice in this extract from our new report, Practical AI for Local Media.


In starting to write this report, I wondered why the Nordic countries are so incredibly well represented in AI applications. The answer, I was told, is that they have a lot of publicly accessible, well organised data and well organised data is the foundation of practical AI.

Elin Stueland, deputy news editor at Stavanger Aftenblad in Norway put it much more simply: “Crap in, crap out. If you have good data, that just solves everything. But if the data is chaotic, then you will never succeed.”

For some of the use cases mentioned in this report, data is sourced from commercial providers. In others, it is taken straight from local and national government sources. In the case of Stavanger Aftenblad’s junior sports coverage, it comes from the football league administration, with match reporting supplied by team coaches using a mobile app.

The one thing all the AI applications we looked at for this report have in common – property prices, sports reporting, weather updates – is that they are all founded on robust, reliable, structured data sources.

“If you have good data, that just solves everything. But if the data is chaotic, then you will never succeed.”

Elin Stueland, Online Editor, Stavanger Aftenblad


The other commonality for local media using practical AI is the sheer scale of the datasets it processes.

The ability to handle massive volumes of data is a key value driver in terms of ROI, in relation to both the volume of data processed and content output. It takes a content robot the same amount of time to create one, one hundred or one thousand articles. The more data you throw at it, the more content it can produce.

Traditionally, publishers have had to decide how narrow to go with their content before the limited scale of audience interest renders reporting commercially unsustainable. But with robots working on comprehensive local datasets, the level of granularity that can be served is potentially limitless.

Companies can use AI to filter property sales data to create top-10 lists of the most expensive houses in a broad metropolitan area. But they can also drill down into property sales for specific neighbourhoods and target that content to drive engagement hyperlocally.

Charlie Beckett, head of the Journalism AI project, listed a range of projects initiated by Journalism AI fellows looking to understand or add value to existing content, from fact checking to adding ‘context cards’ to news articles. He said, “That’s all about volume and, frankly, because a lot of the volume is being created by AI you need, as a newsroom, AI tools to be able to filter it and identify what’s interesting.”


It may be helpful to view a practical AI setup as similar to a commercial newswire service, but working exclusively in-house. The AI creates content automatically according to a set of rules determined by the publisher and the data available with the content created and distributed throughout established publishing systems.

The big difference between the AI and a commercial wire service is that output can be entirely automated to suit specific publisher goals, including subject matter, style, formatting and distribution.

For some publishers, the AI puts property sales information or junior sports reports directly onto the front page of their websites. For Dutch publisher NRC Handelsblad, AI automation is used to pull web stories into personalised newsletters.

The AI uses subscribers’ behavioural data to automate creation and delivery of its midday newsletter. In a lineup of 12 stories, three will be trending that day, while the remaining nine are chosen based on individual reading histories.

As well as saving time and resources on newsletter creation and distribution, the introduction of personalisation has improved newsletter engagement. “We saw that people receiving this email were more active on a weekly basis on our website or our app,” explained NRC’s data and innovation manager Luuk Willekens.

“You can develop products by finding out what kind of robotics actually work for your audience.”

Jens Pettersson, Head of Editorial Development, NTM


With enough training, AI will take raw data and produce individual texts that can be automatically distributed through established publishing systems. But it can also be used as the starting point for broader content packages and to alert reporters to opportunities for a good old-fashioned investigation.

Everyone loves a list and rather than only publish articles on individual property sales, Swedish news group NTM uses AI to compile articles on the biggest deals, region by region. Jens Pettersson, head of engagement & loyalty at the 100 year old publisher told us, “We do this week’s highest value property sale and this month’s most expensive properties. You can develop products by finding out what kind of robotics actually works for your audience.”

We heard that AI content from PA Media’s RADAR newswire is often the foundation for local follow ups and Pettersson said some of the content robots they use at NTM alert the human staff to interesting developments. “They are almost worth more as an alert,” he said. ‘Hey, this house has been sold four times in six months and is getting more expensive every time. What’s going on here?”

This article is an extract from our new report, Practical AI for Publishers, sponsored by United Robots. Download it for free below, and listen to our special podcast documentary episode featuring the voices and experiences of the report’s interviewees here.

Your details will be used to send you the Practical AI report, as well as information about future Media Voices reports and United Robots communications. Please note United Robots and Media Voices are joint data controllers for this report.


Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *