Skip to main content

Kennedy School Review

Topic / Media

AI in the Newsroom: How AI Could Improve the Work of Journalists

I spent the last few days in Perugia, Italy, where I attended the 2023 International Journalism Festival. Such events are not places where journalism happens. They are aberrations, but they let journalists unwind, exchange views openly, and think big.

One of the big issues discussed at the festival was AI. Advancements in AI technology are a threat as well as an opportunity for journalists across the globe. Journalism at its core remains the same: it is about holding those in power to account. But in the AI age, it is going to be about investigating governments that will try to use AI to spread even more disinformation as well as holding the code and the people behind it to account too.

Access to information will not be a problem. We are already living in the age of information overload. The key premium skill (and not only in journalism) will be the ability to establish the truth based on verified facts and scientific evidence. It matters when taking into account the tectonic shift between ad supported business models towards reader revenue models (subscriptions, donations, memberships etc.). Readers will pay only for quality and exclusive information. The New York Times understood it years ago when they pioneered the paywall.

In this context, another trend in corporate storytelling is visible. It is not anymore just about a newsletter. It is becoming more about quality storytelling with journalistic DNA. Organizations are looking for quality, be it a podcast or a newsletter. Raju Narisetti was a career journalist who is now the global publishing director at McKinsey, and I was surprised by his statement that McKinsey’s newsletters have subscribers in the millions. It is a scale not imagined by many media organizations. These are spaces where journalists should look at and be inspired by as well.

Narisetti brought to light an interesting example. Everyone is talking about AI now. McKinsey cannot compete with all the news, but if you google “What is generative AI?” the top result will be a McKinsey article! The conclusion is that good content is not the only thing you need to succeed. What you also need is the experience layer (build your expertise, build habits, persistence advised, aim long term). For some reason, Google’s algorithms recognize the article as a reliable piece of information and rank it high on its results page. Supposedly, there is a lot of authority behind McKinsey’s quality content offer.

Will AI replace journalists? Not that easy. It will replace “content creators” and will free journalists from mundane and repetitive tasks. Specifically, investigative journalism and quality opinion writing will not be automated, but it can be supported by AI in the research process.

Spotlight is one of the best movies about journalism. In the film, Marty Baron says: “we spend most of our time stumbling around the dark. Suddenly, a light gets turned on, and there’s a fair share of blame to go around.” Investigations at their core are not flashy. It is mundane work. It is like putting a puzzle together with many dead ends to misguide you. AI can speed up the process, but the workflow remains the same.

Over a year ago, I was training an independent Hungarian media outlet specializing in investigations; I introduced them to Google’s AI tool for journalists called Pinpoint, which analyzes large groups of documents. They had just started working on a story where they had to process over 1,700 files. Once they uploaded them to Pinpoint, they were immediately able to speed up the research process; the tool started searching for regularities, such as people, companies, and addresses. Most importantly, they were able to search the scanned documents using keywords and work in a team. Eventually they published a series of articles.

AI itself will not solve journalistic problems. It can help, but it should be embedded with values and true journalistic thinking. Even when employing AI, you will get nowhere without knowing where you are heading. To get a story, you sometimes will still need traditional shoe leather reporting skills with a pinch of design thinking.

A few years ago, I was attending an event at the Institute of Politics at the Harvard Kennedy School. One of the panelists was Jane Mayer, a renowned New Yorker journalist, and she said such a brilliant thing: “I was doing a piece on Mike Pence. He never agreed to talk to me, so I went to his hometown and had coffee with his mum. That way I heard some great stories.”

There are already some outstanding initiatives such as London School of Economics’s JournalismAI project, researching the use of AI for quality reporting (supported by Google News Initiative). Platforms are also advancing their AI tools for journalists. It might be a part of their lobbying efforts to placate regulators, but the bottom line is that some of the tools are truly impressive and useful for journalists. Google News Initiative was also the main sponsor of the Perugia conference (disclaimer: I worked as Google News Lab’s main training person for the media in Central & Eastern Europe, but I went privately to Perugia paying for myself).

Speaking about AI means also discussing platforms because they are the wielders of the algorithms. For the media, cooperation with platforms has always been a double-edged sword. The issue was thoroughly discussed during the “The elephant in the room: could AI give technology giants more control over the news?” session. If the media get too close, then they risk dependency on an external infrastructure they do not quite know and control. The dilemma might be how much of their data are they willing to trade off to train an external algorithm. Will it be good for them in the long run?

In addition, platforms such as Google and Meta are not organizations devoted to journalism but rather for-profit advertising vehicles. Do they truly care about journalism, or do they just need us to advance their lobbying efforts? There are more questions than answers here, but it is always worth it to keep the conversation going. It is an adaptive challenge. At this stage, we really do not know where it will take us.

Professor Charlie Beckett, a former journalist now leading the JournalismAI initiative at LSE, stressed that when traveling the Global South he realized an opposite issue there. For many newsrooms, their problem is plummeting revenue and trying to get their audience to deal with political interference. These newsrooms are seeking help in understanding and accessing some of the AI technology, so they do not fall behind. They are aware that if they don’t strengthen their organizations, then other people will start to provide information and control the information sphere.

Journalism will survive the impact of AI, but the future belongs to small and agile media organizations that are data and AI savvy with traditional journalism skills mastered as well, such as Ukraine’s digital native The Kyiv Independent. The ad supported business model for news is broken. The future is about reader revenue, but it involves mastering engagement and community building skills, employing AI models on the business side (e.g. dynamic paywalls). These are things that big corporate media, which are used to ad money and one way relationships with their audiences, have been struggling with. In addition, the scale of reader revenue will never match the good old times. Just a few big players, with the mission at heart, understood it early and started transforming themselves accordingly, such as The New York Times, Financial Times and The Guardian.

Photo credit: “Ashni” via Unsplash