Greetings. I know that you come here to get examples of practical use cases of GenAI in the news media business, so today’s newsletter pulls together a number of these from different corners of the world.

Using GenAI to create money-making content

We have heard about using GenAI to summarise and create different versions of an article. Here’s a news organisation that repackages several articles into one — and then sells that to an entirely new market.

This is the story of Nikkei, the Japanese publishing giant that produces 1,000 articles a day. Nikkei discovered a segment of its audience was interested in its news but not willing to pay the subscription fee of US$30 per month. These readers also found the average article on Nikkei.com difficult to read and too long and time-consuming.

The solution was a condensed news product called Minutes by Nikkei, which took six months to build. It uses generative AI and proprietary algorithms to combine several articles into one, rewritten in an easy-to-understand style that highlights only the main points of the news.

It produces three such articles a day, which are then packaged as Minutes by Nikkei and cost US$7 a month. They are also produced as audio. 

It is editors who select the initial pool of articles the GenAI works from, and editors also review the final product before it is published. “A close relationship between the editors and the development team is essential,” Seiya Shinhashi, the product manager in charge of Minutes, recently told the inaugural INMA Generative AI master class. 

Nikkei had been concerned the new product could cannibalise its premium product, but this did not turn out to be the case because regular business readers were not satisfied with just three articles, Shinhashi said. 

Master class attendees also heard about another case where GenAI is creating content that is being monetised effectively.

Stuff, New Zealand’s largest media company, was grappling with a time-consuming manual process for pitching sponsored content and the high costs of paying freelancers.

Why not try automated content creation? 

“We worked very closely with editorial to make sure that our ethics and principles within the commercial side of the business were very much founded and aligned with editorial principles in the newsroom,” said Matt Headland, managing director of brand connections at Stuff.

Automated content appears on topics such as culture, motoring and real estate, which he said was “a huge vertical for us.”

There is still human interaction throughout the process, with staff editing the content, but AI tools speed the entire process up. In the month of March, Stuff saved 60 hours by using AI.

“Even though it has driven efficiencies within the team, we can very much scale up our revenue because we’ve got more output,” Headland said.

It was initially hard to get staff to adopt the tools because they were concerned about technology taking over their roles, Headland said: “It did take some time for people to embrace the tools that we were putting in place. But once the team saw the benefit and the scale of the content, and also the revenue that it produced, people really jumped on board.

“It was nothing but successful.”

Stuff is also using GenAI to create presentations for “a couple of hundred people” working on sales teams, cutting the amount of time taken to produce a presentation to an hour from a week. 

“We are now seeing revenue growth from this as well because our sales team is now able to spend more time talking to their customers and solving their problems rather than sitting and creating presentations.”

Automated video transcripts, radio traffic reports, and more

What are news brands doing with GenAI right now? 

Well, if you recently came back from five days in London for the INMA World Congress of News Media and study tour, your head is no doubt buzzing with ideas and opportunities.

Here’s a look at some of the GenAI aspects of the event:

At Thomson Reuters, AI is seen as a tool to unlock the human potential of journalism — and every staff member is expected to be alert to the possibilities.

The news agency has created a GenAI video tool that produces a time-coded transcript, a translation, and a scene list. Clients can now search within the video for key terms to quickly jump to the right spot. It analyses scenes, frame by frame — much faster than humans do — and identifies key people using facial recognition technology. 

Reuters is also working on training algorithms with technology companies and corporations. The company is looking at licensing opportunities but also at training a model so it can create headlines or content in the voice of an external organisation using Reuters content. Reuters data will also be used to validate content in RAG frameworks. 

At New Zealand Media and Entertainment, GenAI holds opportunities for revenue, CEO Michael Boggs said. 

“We make a lot of money from radio traffic reports,” he told conference attendees, pointing out that soon, these traffic reports would be generated by AI, producing 4,500 minutes a week of content. Next up: weather reports. And then, news, using a human-trained voice to which NZME will own the intellectual property. Also on the roadmap: uses in advertising.

The use of GenAI for formulaic, commodity reporting (like traffic or financial data) actually raises the value of human journalism, Sky News Group Executive Chairman David Rhodes pointed out: “Equity prices and weather are devalued. The value of what the bot cannot tell you is that much greater.”

Humans will seek out human-created and human-curated content, Hearst UK CEO Katie Vanneck-Smith said, pointing to Spotify playlists versus Spotify’s algorithmic selection as an example. The marketing team at Hearst has been using AI to drive subscriptions growth for a while, but the real potential lies in using it to increase productivity and efficiency at magazines, she said. 

How much of a game changer is generative AI?

The limiting factor to innovation is no longer money or tech resources — it is news organisations’ ambition and imagination when it comes to using GenAI, said David Caswell, founder of StoryFlow Ltd. 

He cited Cuestion Publica in Colombia, an investigative news organisation of about 15 people that built a database of relationships in Colombian society it queries when news breaks to see how various people in the story are connected, instantly providing contextualisation of breaking news.

Cuestion Publica further fine-tuned a model to adopt its sassy voice. The automation produces tweets for X within 15 minutes, compared with the three hours it used to take journalists earlier. 

It will be increasingly important for news brands to listen to their audiences and “viscerally” understand them because knowing what they care about is going to become more and more valuable, Caswell said. 

“Think about things that you can only do with these tools — whole new things that you could never do manually,” he said.

Date for the calendar

Wednesday, June 12: How are news brands using GenAI to help their newsrooms be more productive? Please join us for a Webinar featuring two deeply respected brands — Sweden’s Bonnier News and Germany’s Der Spiegel — that are undertaking innovative initiatives. Free to INMA members. 

Worthwhile links 

• ICYMI: Google and OpenAI are both launching supercharged assistants. Google also made it clear its search generative experience is here and rolling full steam ahead.
• GenAI in advertising: Here’s a look at what six publishers are thinking.
• More GenAI summaries: This time, at Gannett.
• Tips for using GenAI for search:Provide context in prompts. Note that the machines are getting smarter and that prompt engineering may soon be a thing of the past.
• Budgeting for GenAI: Last year, much of enterprise genAI spending came from “innovation” budgets and other one-time pools. In 2024, however, many leaders are reallocating that spending to more permanent software line items.
• 101 use cases of GenAI:Most of them are applicable to the media industry. 
• GenAI for productivity vs products: Employee productivity is driving the first wave of enterprise generative AI adoption with customer service chatbots for low-risk customer use cases being the notable exception. 
• GenAI and consumers: Prepare to get manipulated by emotionally expressive chatbots. Hmm — could this help us sell subscriptions?
• GenAI licensing deal: Reddit signs with OpenAI (after also signing with Google; good for them).
• GenAI in elections:This publication is tracking the good, the bad, and the ugly worldwide.

A non-AI diversion

How is it that some people are just lucky (I consider myself among them)? Three thoughts on creating your own serendipity and six more tips here.

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.

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