When INMA rolled out its Generative AI Initiative in 2024, many questions swirled around the new and unfamiliar technology. The initiative, led by Sonali Verma, emerged to explore and share best practices for using GenAI across the news industry.

During a year-end Town Hall session, Verma observed how greatly the landscape has changed in just two years. AI is no longer a futuristic concept, she noted, but a present-day imperative: “Everyone is using AI. It’s no longer a nice-to-have. It’s very much a must-have if you want to compete against everyone else in the news business.”

Early questions over ROI, build vs. buy, and the trade-off between quick wins and long-term impact have come into focus for news industry leaders, and now a significant focus is on what the new AI-driven news ecosystem means for business models.

Over the past two years, the Generative AI Initiative has connected with more than 150 publishers worldwide, spotlighting their experiments, breakthroughs, and lessons learned.

During her town hall presentation, Verma showcased an impressive range of AI applications, from routine automation to Pulitzer-worthy investigations, to illustrate the many ways it is being used.

Amongst the highlights she shared of how media companies are putting GenAI to work were:

1. Automating routine tasks

Metadata and SEO tagging were among the earliest use cases for GenAI.

“Metadata tagging, SEO tagging, that was everyone’s first use case,” Verma said, noting it is work that “humans don’t love doing and also humans don’t necessarily do well.”

Machines, on the other hand, excel at consistency, especially with messy keyword variations like “COVID-19” in all caps, hyphenated, or lowercase.

Now, automated content placement has become standard, and Verma noted that most brands see a lift of about 5%-20% in total revenue when they do so.

At The Hindu in India, personalised homepages, article recommendations, and push notifications led to a 4x increase in app pageviews and 40% of subscriptions happening in-app. The Times of India uses AI to rewrite headlines based on audience segments, which enhances engagement.

2. Automating audio

Transcription tools are also now part of the reporter’s toolkit, but some companies are taking it further. To give its audience more live coverage of football matches, the BBC piloted a tool that transcribes audio commentaries and delivers quotes and summaries based on those transcripts.

In Norway, Schibsted uses AI-generated audio playlists to serve younger audiences and boost completion rates. And Time Magazine and Business Insider offer personalised audio briefings, allowing users to choose voice type and summary length.

3. Supercharging reporting

AI has proven to be a powerful tool for improving reporting, and Verma said it has even helped with Pulitzer Prize-winning investigations.

In the first case, the Associated Press used AI to transcribe hundreds of hours of police bodycam footage and extract data from 200,000+ documents in a three-year investigation into police-related deaths.

In a second example, the Washington Post deployed object detection models to analyse drone footage in a forensic investigation of journalist deaths in Gaza.

“The visual forensic journalism team was looking into the Israeli military’s explanation for why two journalists were killed in Gaza, and so they obtained and reviewed drone footage,” she said. “They found that no Israeli soldiers’ aircraft or military equipment of any sort were visible in the footage that was taken that day, which raised a lot of questions about why those journalists were targeted.”

The Wall Street Journal also used AI for an important story, analysing 41,000 Elon Musk tweets to track his political rhetoric. And a collaborative project between the Center for Public Integrity, Reveal, and Mother Jones used image recognition to uncover post-Civil War land dispossession among formerly enslaved people, tracing descendants through AI-generated family trees.

“In each of these four cases, this is a lot of work and AI really helped them get through it relatively quickly,” Verma said.

4. Data and feedback loops

Behind the scenes, AI automation enables the creation of dashboards that provide greater insight into customer needs.

Schibsted Norway built a pipeline to analyse customer feedback, tracking sentiment and urgency. The result is impressive: 70% more feedback captured and analysis time reduced from days to minutes. Similarly, the Telegraph in the UK replaced multi-page reports with instant AI-powered data queries, benefiting the entire organisation.

5. Customer service and voice bots

AI is an effective stand-in for humans when solving specific problems. Medienhub in Germany deployed a voice bot to handle issues with newspaper delivery. It authenticates callers, provides updates, and passes context to human agents when needed. Only 5% of users request a live agent, and the project paid for itself in six months.

6. Sales and advertising

More applications are being discovered for using AI in sales and advertising; at Hearst Newspapers, ad sales reps work with an AI assistant that prepares personalised pitches and interprets data — built by a non-technical sales rep.

The New York Times developed Brand Match, which scans content to create audience segments for ad campaigns. It outperforms traditional campaigns by 32% and industry benchmarks by 62%. And Meredith (now People Inc.) uses AI to infer user intent from content consumption, identifying unexpected correlations like family travel readers also engaging with Nintendo Switch content.

Other uses for AI include summarisation and accessibility, translations to reach multilingual audiences, and rewriting content geared for a specific audience. For example, Frankfurter Allgemeine Zeitung in Germany introduced bullet-point summaries to help time-strapped readers — and was surprised to find this helped boost subscription conversions.

In Japan, Nikkei launched a $7/month product offering that simplified, AI-generated versions of its top three daily articles, targeting readers who found its traditional $30/month offering too dense or expensive. And the Times of India rewrites news for children with a friendly tone, trivia, and audiovisual elements.

Looking at what’s next

As Verma noted, the industry has moved from AI experimentation to scaling. The focus now is on adoption — getting entire newsrooms and enterprises on board. But she cautioned that using AI solely for efficiency yields only incremental gains.

“Productivity does not equal value,” she said. “Time saved is not always money saved.

“It’s not until you actually use the new time that you’ve got for either a revenue-producing activity or to fundamentally change a business process across multiple employees across the entire system that you’re actually going to see financial savings.”

To truly unlock AI’s potential, news organisations must shift from table-stakes automation to strategic differentiation, Verma emphasised.

From metadata tagging to multilingual monetisation, AI is reshaping journalism’s possibilities. But the real opportunity lies ahead: using AI not just to keep pace, but to leap forward. AI threatens the news media industry in many ways, but learning how to use it properly can make operations effective and efficient.

“We can do it,” Verma said. “We can use it to do things that we were never able to do before, like reach new audiences in different languages and formats and monetise them more effectively.”

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .