Want to know what your peers are doing with generative AI? This edition features a smorgasbord of use cases for you to feast upon and be inspired by. 

We look first at the broader trend of GenAI use cases, including 10 interesting examples across the news media industry, three of them with a particular focus on ensuring the risks associated with GenAI are kept in mind. 

And then we turn to a trend that has been on the rise in recent months and that appears to be paying dividends for media companies: audio.

Trends in GenAI across the news business

I went through the submissions for the INMA Global Media Awards, where there are 52 different use cases from news publishers using AI. All but 10 of them use generative AI. Of the 42 remaining entries, here’s a look at how they can be categorised:

• Editorial: The submissions overwhelmingly targeted editorial purposes: 33 of the entries relate to newsroom products or functions. 

• 19 were for making internal workflows more efficient or effective.

• 14 were for creating external, customer-facing products.

• Customer service: Three were related to smarter ways of undertaking customer service — either to learn about customers or to interact with customers.

• Advertising: Three targeted advertising.

• Marketing: Three were for marketing campaigns

As you can see, most news publishers are gravitating towards ways to help their journalists do their work more easily. What’s interesting is that there is no cutting of corners here: You can see the finalists are particularly thoughtful about addressing the risks that accompany AI.

For example:

• Rather than letting deep-seated cultural fears derail their innovation process, Omni focused on the importance of getting journalists on board by letting them design the GenAI tools.

• Ippen Digital even invited journalists to spend two weeks on its AI team co-creating the tools.

• Hearst Newspapers built in a peer-review process — deliberately creating friction with its AI tool to ensure journalists think carefully about how they are using it rather than just automatically ticking boxes.

(Note: I am not involved in judging these awards, so these comments are not an indication of who is going to win. We’ll both find out who wins on April 25 in London at the INMA World Congress of News Media.) 

Need more inspiration? Here are some cool projects that are underway as part of the JournalismAI initiative. The dominant theme appears to be using GenAI for extracting stories from data for investigations. Also: building chatbots.

1. AURA: Advanced Understanding and Research Assistant: a conversational AI platform designed to help journalists navigate and extract stories from complex unstructured data, offering instant context and investigative leads.
2. Know Your Leader: An Al-powered chatbot that lets the public track political speech, comments, and opinions. 
3. IntelliNewsComparer: Uses machine learning to semantically compare text documents in English, Finnish, and Tagalog via pre-trained LLMs and a user-friendly interface for querying document similarities. It can expedite the examination of documents in investigative newsroom projects.
4. CheckMate: A simple Web app for real-time fact-checking on live or recorded video and audio broadcasts to identify and debunk false claims and enable newsrooms to actively work to prevent the spread of misinformation over a significant election year.
5. Real Estate Alerter: Harnessing anomaly detection methods and LLMs to uncover hidden news stories within real estate data, creating an alert system, which provides reporters with a head start on newsworthy stories.
6. StyleCheck: An AI-driven tool to check adherence of articles with the newsroom’s style guide, integrated with a cleanup tool for article copy.
7. Data Robot AIde: An open framework for data-driven and LLM-powered AI chatbots that helps newsrooms create chatbots using different structured datasets to speed up storytelling and help find story ideas. 

You can read more about them here. 

ROI on GenAI: the potential of audio 

Publishers are becoming far more ambitious and innovative about using GenAI in audio than they had been in the past — and finding the ROI on audio-related tools often makes them worth investing in. 

Take, for example, the INMA Global Media Award submission by Germany’s Medien Hub Bremen-Nordwest, a joint venture between regional newspapers Nordwest-Zeitung and Weser-Kurier. They were dealing with print subscribers phoning them about delays in newspaper delivery. 

It is a situation that will sound familiar to many of us: Their agents were often dealing with long queues, and their customers were often dealing with long wait times. The biggest concern? Cancellations because of poor customer service.

Medien Hub wanted to relieve its agents of 20% of the cases and increase customer service availability significantly, especially at peak times. It worked with an AI tool, Parloa, to build a voicebot. Within six weeks, the bot ended up processing 30% of telephone complaints and significantly improved customer service availability. 

Even more impressive: The AI voicebot paid for itself within six months — which is half the time that Medien Hub had expected it to take.

“Our minimum viable product (MVP) goal was to achieve a return on investment on project costs plus ongoing costs within 12 months. However, the bot proved to be much more productive than expected from day one, leading to the ROI being reached in just six months,” Fabian Rosekeit, head of CRM and growth at Medien Hub, told me.

(You can hear more about it from Rosekeit himself and ask him your questions at our Generative AI Master Class in May.)

Medien Hub has now expanded the AI voicebot’s service to include vacation holds as well and is looking into letting customers submit complaints about past service, rather than simply current issues.

Another audio use case that generates efficiencies is audio transcription. Many news publishers have told me they are building these tools; Norway’s Schibsted has developed a tool called JoJo (using OpenAI’s Whisper model), which has already saved its journalists more than 18,000 hours of work over the course of a year. Other publishers are also using JoJo now.

Switzerland-based Ringier AG’s Blick built a tool that not only transcribes audio and video content and translates it into Swiss German but also exports subtitles as needed. The time required for transcribing was reduced from up to four hours to mere minutes — and journalists can now upload files immediately after interviews, with transcriptions ready by the time they return to their desks, thus speeding up the publication process.

As far as consumer-facing applications go, Spain’s Prisa Media uses audio for its personalised AI voice assistant Victoria (built in collaboration with Amazon, it works on the Alexa product). Victoria lets football fans pick their favourite team and ask questions about it. This allows the audience to engage with and interact with content on Prisa’s radio stations in a new way.

A Gazeta in Brazil clones its reporters’ voices so that they need to simply submit text to create audio voice-overs for videos.

Is consumer-facing audio monetisable? Well, The New York Times’ subscription-only app just passed a million downloads in about seven months.

How about GenAI audio as a way to build trust or to reach new audiences? It could be a play for younger consumers or for busy people who are multitasking — e.g. listening while cooking or driving their kid to sports practice or out for a run. 

But Dutch public broadcaster NPO is looking at an entirely new, even surprising, end user for audio: a listener who has trouble hearing. 

“We are experimenting with visual enhancements of podcasts,” said Ezra Eeman, director of strategy and innovation at NPO. “You can not only experience a podcast as something you listen to, but you can also visually experience it.” 

In other words, automatically creating video content from audio content.

Eeman also envisions a future in which instead of listening to a podcast, you could have a conversation with a podcast — a natural extension of the text chatbots many publishers are already building.

ICYMI: For a deeper dive into audio trends, take a look at INMA’s recent report, Why Some Media Companies Are Betting Big on Audio.

What I’m hearing

The 80-20 rule: No, not the Pareto Principle but the GenAI edition of it, which has two variations:

• Get AI to do 80% of the work and humans to do 20% (always keep a human in the loop)(h/t Tim O’Rourke and others).

• Buy 80% of the GenAI tools you use, and build only 20% — customise them to suit your needs instead of reinventing the wheel at great expense.

Date for the calendar

Wednesday, April 3: The Financial Times has just launched its first reader-facing generative AI product to address the challenge of meeting the audience where they are headed. Please join Chief Product Officer Lindsey Jayne on a Webinar as she pulls back the curtain on the FT’s approach to GenAI. The Webinar also features Eduardo Lindenberg, innovation director at Rede Gazeta, who will talk about an acclaimed chatbot that helps the news brand reach younger audiences more effectively.

Worthwhile links

• Future-proof your news business in the GenAI era: The Wall Street Journal and Hearst Newspapers took the stage to tell us how — and it was incredibly reassuring and insightful.
• GenAI tools: This space is moving fast. Andreesen Horowitz ranks top GenAI tools and notes with surprise that over 40% of the companies on the list are new since its September 2023 report.
• GenAI for customer service: Strong ROI numbers in the financial services industry, as well as the ability to remind customers about what you want to nudge them towards.
• Beware the black box: LLMs sound really confident about what they know, but that can be misleading.
• Beware the black box II: “Unfortunately, the web isn’t really a trustworthy place.” 
• GenAI licensing deals: What is your news worth to a large language model?  

Diversion

You know how AI agents are going to do all the work for us in the future? Well … if you’re too busy to swipe Tinder or Bumble, here are the green shoots of that future.

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.

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