AI in news: Key takeaways from the International Journalism Festival

Photo: Ilaria Sofia Arcangeli

When I first spoke about AI and news at the International Journalism Festival in 2018, the room was quite small and some seats were empty.

Last month, I returned to the Festival in the beautiful Italian city of Perugia to chair another talk on AI and automation. Though the room was much bigger, some people were still turned away due to lack of space.

At this year's IJF, there were several sessions on AI and journalism – all of them with long queues and more people than chairs. The release of ChatGPT last November has changed the dynamic. Suddenly, newsroom AI is not a minority interest of the geeks in the corner, but a looming question for news organisations and individual journalists alike.

The role of artificial intelligence in journalism was one of the big themes of the festival – not just in the sessions but also in countless conversations taking place in the corridors and cafés. There is real excitement, plus a good deal of concern, among many news professionals as to how to respond. So, where are we on this fundamental issue affecting the future of journalism?

It's not just hype

David Caswell is an executive product manager at BBC News and a long-time innovator in the field of automation and AI. During our IJF session, I asked him to characterise the scale of disruption generative AI could bring to the news industry. “I think this is going to change journalism more in the next three years than journalism has changed in the last 30 years,” he said.

As someone who has worked in newsrooms for the last 30 years, I find David's comments slightly concerning. As an industry, our track record in dealing with fundamental change has not been good. News organisations failed to grasp the level of disruption bought by the arrival of the internet in the mid-90s, and did little better when social media appeared 10 years or so later.

Are we going to do better this time? There are some promising signs: some news companies have grasped that this is a disrupt or be disrupted moment, and have already got to work.

A genuine moment of change

At the Guardian, head of editorial innovation Chris Moran recognised the release of ChatGPT as a significant moment of change. He said that it made him stop worrying about the complexity of the technology, and instead roll up his sleeves and play.

All around the Festival, I heard about people experimenting with the technology for everything from news discovery within scientific research papers, to writing headlines or SEO keywords, to proofreading articles, summarising them or even creating different versions.

Over at the German public broadcaster Bayerischer Rundfunk, head of the AI and automation lab Uli Köppen said she and her team were busy testing and prototyping – but not yet using ChatGPT in anything audience-facing.

There were words of caution, or perhaps experience, from Lisa Gibbs, head of news partnerships at the Associated Press. “The rush to sound smart and try things is leading newsrooms to rush to do some things that they don’t have an understanding of," she warned, "but more importantly their newsrooms do not have the infrastructure to support.”

This is an important point that was picked up by David: AI might enable newsrooms to produce multiple versions of every piece of content, but what then? “You need infrastructure to be able to use the correct version for the correct audience in a way that is valuable to audiences,” he said. “That’s not just the content management system, but the way you curate that stuff and recommend it.”

While it is exciting to experiment, doing anything at scale presents a different set of problems and solving them is big, strategic, and potentially costly. Organisations need to think hard about how the opportunities with AI fit with their USP, their values, and their ambitions.

The impact on individual journalists

Most of the AI sessions in Perugia heard similar questions about AI and job losses. Semafor’s Gina Chua was forthright on this: “We are not here to save journalists. We are here to save journalism.”

Meanwhile, Lisa Gibbs pointed out that journalism was already losing thousands of jobs, not because of AI but because of the economics of the news industry.

My own view? Some jobs will go, others will be created, but the biggest change will be in the day-to-day tasks carried out by journalists. As Jessica Davis, director of AI and automation at Gannett, said: “I am seeing this as a cultural change. How can it be an additive to our journalism? How can it free up our journalists?”

AI tools will doubtless remove or reduce the need for some mundane jobs. They will transcribe interviews and write repetitive, highly structured stories. In a recent podcast, Matt Karolian from the Boston Globe suggested that AI would give journalists more time to report as they would be "untethered from the chore of writing".

Most journalists I know enjoy "the chore of writing" and would be loath to give it up, but maybe soon they’ll have an AI sub-editor alongside them to help. Journalists will need to focus on tasks where they are more effective than machines and look at how their work can be enhanced by adopting AI tools.

Newsrooms will also need translators who speak both the language of editorial and tech. Get familiar with the technology. Become one of those translators.

The opportunities

In Perugia, there was optimism about the use of AI tools to take the friction out of workflows and make newsrooms more efficient. This feels like step one in finding a good use case for AI. Step two is to enhance what you do already.

At BR, Uli said they were looking at embedding natural language generation in the core of their newsroom systems. "So, for example, when you get your agency report you would have an automated summary and different headlines that are suggested," she explained. "When you are writing about something, different articles are suggested to the author from your system, so they don’t have to do the research.”

The third step in deploying AI tools is – for me, anyway – the most exciting: enabling newsrooms to do things that are not possible by human effort alone.

We have already seen examples of this in the use of AI to sift large amounts of documents, finding information vital to investigations. What else might we be able to deliver to audiences? Better engagement, new forms of storytelling, news customised to the needs of each user? Gina Chua pointed out that to get this right, what we need in newsrooms is imagination.

What to do next?

For any news organisation, this is a difficult moment: do you get cracking with AI, or stand back from the hype and see where it goes? Jessica aptly described the Gannett approach as "cautious, but curious".

Meanwhile, the BBC would be a "fast follower rather than initiator", David predicted, noting that at a time when the corporation was doubling down on trust as its key value for audiences, exciting new uses of generative AI may not be best.

However, many people at the Festival seemed to agree on the need to make a start, at least, in looking at what generative AI – and the wider range of automation tools – could offer. Alongside that, the need to consider how AI fits your organisation's values, and to start building some guidelines to give your teams the confidence the use it.

Ultimately, it's worth remembering that this is not just about technology, it's about journalism and the audiences you are seeking to serve. So take a deep breath, think hard, and start small.

Gary Rogers is a Senior Newsroom Strategy Consultant with Fathm. In 2018, he co-founded RADAR, an AI-driven local news agency operated in the UK by PA Media.

Contact gary@fathm.co if you are seeking support in your strategic approach or implementation of AI.