GenAI resolutions for 2025 + interesting use cases

By Sonali Verma

INMA

Toronto, Ontario, Canada

Connect      

Happy New Year, everyone! A fresh beginning to a year in which I’m sure many of us will achieve great things and will learn a great deal. 

I thought it would be nice to start off the year with some positive, inspiring examples of how AI is actually helping news media businesses serve their staff and their audiences better.

Also, what is a new year without resolutions? Please indulge me while I share some best practices with you that you may want to adopt.

I look forward to hearing about your experiments, your learnings, and your successes. Please do drop me a line to let me know what you’re working on. I am always looking for new voices to feature on our GenAI Webinars.

Sonali

Resolutions for using AI in 2025

• Resolve to use GenAI to make money. The Reuters Institute for the Study of Journalism’s annual predictions report, which surveyed 326 news leaders from 51 countries, identified a few focus areas for 2025: audio and video as well as new products around games, education, translations, and youth.

GenAI tools exist to help you create these monetisable products quickly and easily. Resolve to use the technology where it makes the most sense — and to lean heavily on human judgment where that matters.

Video is important for publishers in 2025. Taken from the Reuters Institute report.
Video is important for publishers in 2025. Taken from the Reuters Institute report.

Audio, versioning, and chat products also feature prominently, according to the report.
Audio, versioning, and chat products also feature prominently, according to the report.

• Resolve to always keep a human in the loop. GenAI is not at the stage where you can hand it the keys and hope for the best. 

• Resolve to invest in training people and managing change. Any tech transformation is a people transformation, as McKinsey says, so double down on building skills. Resolve to involve your end users in what you are building. Find your champions. Support them and help them succeed.

• Resolve to be transparent with your staff about using AI. Resolve to build trust. Communicate: Listen to what their pain points are, and try to resolve them. Show them how what you are building is going to be genuinely useful for them. Listen to and act upon their feedback about how it genuinely is not useful.

Training matters because many AI use cases will affect staff, the Reuters Institute survey shows.
Training matters because many AI use cases will affect staff, the Reuters Institute survey shows.

• Resolve to be transparent with your audience about how you are using AI. They are going to find out anyway. Don’t ruin your credibility. Show them, instead, that you can be bold and innovative; tell them about guardrails you are putting in place; and get their feedback on what works and what doesn’t. 

Audiences perceive news labeled as AI-generated as less trustworthy, even when articles themselves are not evaluated as any less accurate or unfair. But “negative effects associated with perceived trustworthiness are largely counteracted when articles disclose the list of sources used to generate the content,” this research paper points out. 

• Resolve to use GenAI to tackle audience disengagement amid news avoidance. It has never been easier to move beyond the 800-word inverted pyramid article as a standard unit of information.

Give people news that is relevant to them, and create it in a format they find useful. Use data to find what they consider relevant. Experiment with different versions and formats to see what works. If you’re feeling stuck, consider using AI for solutions journalism, as Scroll.in does, so your audience gets a break from negative information.

• Resolve to use GenAI to solve real problems and challenges, not just for the sake of keeping up with everyone else (about 87% of the publishers the Reuters Institute surveyed say newsrooms are being fully or somewhat transformed by Gen AI). 

Consider asking these three questions that Upasna Gautam, senior product manager at CNN, asks before implementing AI in the newsroom: What’s your biggest goal? What’s your most significant pain point? How do you measure success? These form the skeleton of your AI use-case framework.

• Resolve to be creative. What does the future of the news experience look like or sound like? 

“As AI moves from search engine to conversation partner, UX patterns are shifting to spaces that morph around user intent rather than forcing users down pre-defined paths,” according to an AI summary of a report by Ezra Eeman, strategy and innovation director at Dutch broadcaster NPO.

Conversational interfaces for news will help gauge user intent. From the Reuters Institute report.
Conversational interfaces for news will help gauge user intent. From the Reuters Institute report.

How will we produce content in ways that gauges or meets that intent? Conversational interfaces are top of mind for many publishers right now.

• Resolve to think about the ROI for use cases that matter. Need some inspiration? Check out my last two reports.

Cool GenAI use cases 

It’s been a few weeks since our last newsletter, and a smattering of interesting GenAI use cases has emerged in the meantime.

• Here’s a neat approach to building trust through transparency with your audience from software company Every, which also writes about technology. It offers readers the option of seeing the source material, including what’s not in the article.

Screenshot of an article from technology company Every.
Screenshot of an article from technology company Every.

• Thinking about audio? The New York Times uses AI to match videos to audio for its Snippets experiment to help its audience discover programming by swiping through and listening to bits of different articles or podcasts. It also uses language models to help select those clips and for automated voices.

Screenshot of the NYT audio app.
Screenshot of the NYT audio app.

• AURA is a tool to help reporters undertake research before they write a story, built by The Economist, Indian Express, DR (Danish Broadcasting Corporation), and Aftonbladet under the JournalismAI initiative. It lets journalists chat with a database to give them context about how big a story is.

They can upload an academic paper or a press release to get the process started. AURA generates storyline ideas as well as a headline and a pitch for editors, all of which can be further refined through prompting to suit the publication’s needs.

Then, a basic research report is generated, which features different personas appropriate for the story that answer questions based on publicly available information. The next step is to get AURA to recommend potential sources for interviews and future storylines as well.

Screenshot of AURA, an AI research tool developed by four news organisations.
Screenshot of AURA, an AI research tool developed by four news organisations.

• Checkmate, a collaboration between Germany’s Deutsche Presse-Agentur, News UK, Mexico’s DataCritica, and the British Broadcasting Corporation, is a real-time fact checker for broadcasts. It provides a transcript for videos and highlights in yellow any claims being made, then searches for sources to verify them and provides links to them. It was built as a tool to help journalists figure out what is accurate and what is not. 

• The Financial Times conducted an investigation into missing Ukrainian children who were being put up for adoption in Russia. It used AI to join tens of thousands of listings from different databases and then to narrow down potential abduction cases. Reporters then worked off this smaller list to confirm identities.

“Ultimately, it illustrates the importance and impact of taking a really big data set and taking algorithmic tools to expand the possibilities of what were able to do on our own and work very quickly,” said FT visual investigative journalist Peter Andringa.

• U.S.-based Hearst and Gannett, Canada’s The Globe and Mail, and E24 in Norway have built a tool that alerts journalists to interesting real-estate events, using clustering techniques for anomaly detection and LLMs to identify newsworthiness. The LLM prompts were designed after real estate reporters defined — and weighted — criteria for newsworthiness. 

GMA Network in the Philippines and Finland’s Helsingin Sanomat have built a document-comparison tool to help journalists quickly compare large sets of documents.

For example, it can be used for comparing political parties’ election platforms against government coalition agreements to see whose messages prevail or analysing stakeholder comments on legislative proposals to identify supporters and opponents along with their reasoning. It works in English as well as the national and regional languages of the Philippines and Finland.

Time Magazine lets you chat with an article, asking questions, both via voice and text. You can also listen to it in five languages or summarise it.

Screenshot of Time Magazine’s AI features.
Screenshot of Time Magazine’s AI features.

The Washington Post is using GenAI to engage its subscribers in the comments section. Some features of its “Conversations” experience include:

    • Prompts: AI-generated questions will appear with most journalism to offer subscribers a starting point for a dialogue.
    • The Washington Post is using GenAI to engage its subscribers in the comments section: Some features of its “Conversations” experience include: Subscribers can immediately understand their peers’ dominant takeaways and reactions to the journalism through AI-generated summaries of the comments.
    • Sentiment reactions: A range of sentiments, including “thoughtful,” “clarifying,” “new to me,” and “provocative,” are available for subscribers to characterise their peers’ comments.
Screenshots from The Washington Post demonstrating its AI-driven commenting features.
Screenshots from The Washington Post demonstrating its AI-driven commenting features.

For the record, its subscribers did not immediately love the feature. There are 995 comments on the blog post about the feature, many of them negative.

Screenshots from the comments on the blog post announcing Washington Post Conversations.
Screenshots from the comments on the blog post announcing Washington Post Conversations.

My opinion?

When you try something new, sometimes your audience loves it and sometimes it doesn’t. Fortune favours the bold. Kudos to The Post for transparently trying something that hasn’t been done before and learning something about their audience in the process. As of writing, the feature was still being tested, and readers were interacting with it.

Worthwhile links

• GenAI and INMA: I get many questions from INMA members on different aspects of GenAI. The start of the year feels like a good moment to remind you that, as part of the INMA community, you have access to many resources on GenAI, such as:

  • Reports on use cases and best practices.

  • The twice-monthly GenAI newsletter, for which you can sign up here.

  • Webinars. Our next one will be on the playbook for GenAI in the news media business, for which you can register here.

  • The GenAI Master Class, attended by about 150 of your peers in 2024. You can sign up for the next one, in April, here. I’m in the process of putting together the agenda for it now.

  • The GenAI seminar at the INMA World Congress of News Media in May. 

• GenAI and summaries: Looks like Apples new summary feature made a mistake with a BBC news alert and generated a false headline. Apple is moving slowly on fixing this.

• GenAI and advertising: Technology is sending shock waves through the marketing world, promising to create and personalise ads cheaper and faster than ever.

• GenAI and training: Thousands of documentaries have been used to train LLMs.

• GenAI and training: Looks like we’ve run out of human-made data?

• GenAI and revenue: Are AI platforms’ revenue-sharing deals a good move for publishers?

• An AI diversion: AI and whisky: Is nothing sacred

• A non-AI diversion: Many North American cities have a tradition of naming their snowplows. For example, depending on where you live in Canada, your way may be cleared by Austin Plowers, Ctrl-salt-delete, Better Call Salt, Darth Blader, Scoop Dogg, or Sled Zeppelin. In that vein, please enjoy the names of the plows in Wichita, Kansas.

About this newsletter

Today’s newsletter is written by Sonali Verma, based in Toronto, and lead for the INMA Generative AI Initiative. Sonali will share research, case studies, and thought leadership on the topic of generative AI and how it relates to all areas of news media.

This newsletter is a public face of the Generative AI Initiative by INMA, outlined here. E-mail Sonali at sonali.verma@inma.org or connect with her on INMA’s Slack channel with thoughts, suggestions, and questions.

About Sonali Verma

By continuing to browse or by clicking “ACCEPT,” you agree to the storing of cookies on your device to enhance your site experience. To learn more about how we use cookies, please see our privacy policy.
x

I ACCEPT