Deciding where to start in identifying AI use cases for publication and practical implementation can be daunting. From securing data sources to getting your colleagues on board, here are some tips from people who are already working on it, excerpted from our Practical AI report for local media.
‘just do it’
Apologies to one high profile sporting goods company, but the best way to get started with practical AI is to “just try it.” Start with a small project, convince your news editors that it will bring real benefits to them and the publication, and get started.
But before you get started, learn some practical AI basics so you can have informed conversations. “I think there is a general misconception about what AI means,” said Joseph Hook, his RADAR editor at PA Media. ‘ said.
The JournalismAI Starter Pack is a great starting point for beginners who want to understand the great potential of AI. The Journalism AI project has developed a free online resource to help those beginning their AI journey start exploring these technologies. Professor Charlie Beckett, Director of the Journalism AI Project, said: Please educate yourself a little. ”
find a mate
It’s also a good idea to involve people who are at least curious about what AI can bring to the organization, but those people may not be in the IT department. said Beckett. “Otherwise, you’re wasting your time creating a very shiny tool that no one wants to use.”
Stavanger Aftenblad’s online editor Erin Stuland also said it was important to bring the newsroom along to help them understand that automated content is not going to replace journalism. agreed. It’s about highlighting the benefits that AI brings. Uncover the stories in your data, provide alerts on unusual activity, and publish updates quickly.
All of this frees up newsroom time for what Cynthia DuBose, McClatchy’s VP of Audience Growth and Content Monetization, calls “true journalism.”
start small
Actionable AI should be built on realistic expectations, not ChatGPT-level hype, and the general advice is to start small. Hook said it will look for small steps to help improve the process. “Let’s automate what we’re already doing,” he suggested.
Beckett, who has worked on dozens of AI projects since the journalism AI project began in 2018, said it’s best to start relatively small before jumping into something more strategic. “Get used to doing 1 or 2 projects for him using 1 or 2 technologies,” he said. “But people scale up very quickly. I’m really surprised that some relatively small or medium-sized news outlets have said, ‘Let’s do it.'”
Cecilia Campbell, chief marketing officer at United Robots, said many clients start with a small area of data covering a few postal codes. This will help you understand what to measure, what is working and what is not. “Usually that’s the model,” she said. “They’re going to do a small, cheap pilot first.”
“If you hear that AI should be used, first understand what AI is, what it entails, and what small steps you can take to improve processes in your workplace. that is important.”
Joseph Hook, RADAR AI Editor
problem, problem, problem
None of the people we spoke to for this report said much about the technology they use. We talked in detail about the problem we were all trying to solve.
NTM’s head of editorial development Jens Pettersson said it’s important to know your audience, their information needs and the gaps that need to be filled. he said: “What are your concerns? What do you need to change? Would you like to?”
With Stavanger Aftenblad’s junior football coverage project successfully reducing dropouts, Stuland is looking at things a little more positively. As long as there is data, the road to possibility will continue. ”
NRC data and innovation manager Luke Willekens said AI should help journalists or readers, ideally both. “We wanted to find a way to get readers more involved in journalism and give editors more time to do other things with automated, personalized newsletters.”
Protect your data sources
Data is the fuel that drives actionable AI in local media. Whether the datasets you use come from local or central governments, sports associations, or commercial data suppliers, it’s important to have a reliable source of accurate data. Remember, Stuland said that if your data is chaotic, you will never succeed.
Phuc said most of the data used by RADAR is public data from public sources such as governments and health services. But they also use data from a handful of charities big enough to have their own localized data, showing how spending habits have changed since lockdown. For this purpose, we used top-level data from payment card providers. “Basically every town saw people coming back to the pubs,” he says.
Similarly, for projects like NRC’s personalized newsletter project, integrating internal audience data with automated systems is critical to delivering meaningful products and supporting ongoing communication with users. .
Plan output and distribution
The end point of your automated content is just as important as the starting point. NTM’s Pettersson said it helps think about where and how to publish automated content.
NTM uses algorithms to give editors control over the front page of their website. In addition to the start and end times, the editor assigns stories a news value from 1 to 6, depending on the importance of the news. Pettersson’s team sets a threshold for property sales, and properties that exceed the threshold are given higher news value, so the algorithm will place them better on the front page.
“Try to be smart about how you visualize this kind of content in your editorial mix,” he said. “So it’s not just about being at the bottom of the page. Use that as real value for your customers.”
Promotion is as important as presentation. Publishers should let people know that this new content is available. “You have to really think, ‘How can I get the most out of it?'” Campbell said. For some reason people think that success is built into transportation and other things. It cannot be left open and should be treated and promoted like any other content. ”
“You have to think hard. How can you maximize what you get out of it? We have to do, measure and drive.”
Cecilia Campbell, Chief Marketing Officer, United Robots
Iterate toward a clear goal
Treating automated content like any other means applying clear KPIs. Regardless of what a real-world AI implementation is designed to achieve, its effectiveness should be measured, from reducing churn to increasing engagement.
Everyone we spoke to described the iterative process they went through to get their automated content out where they wanted it. Stavanger Aftenblad’s Stuland has found an interesting way to report matches without naming specific players and explained the changes that have been made around player protection.
Willekens explained how changing the selection of articles in the newsletter is important because the newsletter contains too many opinion articles, sometimes 8 out of 12 articles. .
DuBose said the promise he made to the editorial team was that he would never publish automated content until he said it was ready. It took a lot of interaction to train the bot to deliver acceptable text. “They edit it, go away, come back and edit it again,” she explained.
“They said, ‘Dude, we’d never describe it like this.’ This is how we write it. They were the decision makers.”
Pettersson advised a “just launch” approach, but was unable to measure the impact. “Put it out there and see how the audience reacts. Follow up regularly. Change it one song at a time and see how your audience reacts, and if they react badly, undo it and try something else.”
For free access to the report, please complete the form below.
Your details will be used to send you hands-on AI reports, as well as information about future Media Voices reports and communications with United Robots. Please note that United Robots and Media Voices are joint data controllers for this report.
Republished with kind permission of Media Voices, each week we feature all the news and opinions from across the media landscape.
By Peter Huston
Peter Huston is co-host of the Media Voices podcast. He is a writer, consultant and trainer committed to helping publishers build a sustainable multi-platform future. He runs his Guardian masterclasses, speaks at Google’s ThinkPublishing conference, and has been a judge for several magazine awards. He has written about media for The New Statesman, The Drum, FIPP, InPublishing, Publishing Executive, and researched and authored “State of the Media, 2017” for The Media Briefing.
