When we talk about innovation and particularly the use of AI, large European companies tend to dominate the conversation. Theirs are often groundbreaking use cases, implemented in structures that allow them to move quickly because they have long embraced digital transformation.

Today’s newsletter is going to be different. We are going to surface use cases from Latin America and from smaller publishers — scrappy, plucky organisations that are not deterred by a lack of resources and inspired by the potential of this technology. Even better, their use cases translate well across geographies and organisational structures.

What is your organisation doing with GenAI? I would love to hear from you and to connect you with your peers in the news media industry. Please get in touch.

AI innovation in Latin America

At Rede Gazeta in Brazil, AI is being used to make it easier for users to find information, and its applications go beyond the newsroom.  

“We are doing small experiments that could be relevant for our company and, mainly, to improve the experience of our users,” said Ronaldo Henrique da Cruz, subscriptions manager.

He described a new project his team is working on: 

“In March of this year, the income tax declaration period begins. Many users have questions about tax declaration, and we have a lot of content in our collection that can help users clarify these doubts.  

“This project has always worked with the support of our call centre, which assisted users and sent questions to specialists, then our agents received the response from the specialist and responded to users — a very long journey. 

“In 2023, we started using a chatbot that searched for articles that we inserted into a platform we use called Zendesk, and the number of users who were able to resolve their doubts with the articles was very high, greatly reducing contact with our agents (but, even so, we had to provide human support for some users). 

“For 2024, we will use AI to find these articles in our collection with more precision, making the search even easier and expanding the recommendation of more articles that may make sense to those doing the searches.” 

Similarly, Pedro Iván Quintana Garcia has found that it is valuable to assess where GenAI can be truly useful in saving time for staff and where to be more skeptical. He is a director at Ovaciones, a small newspaper in Mexico City that is part of a larger organisation, Organización Editorial Mexicana. 

“When I was offered this job six months ago, I saw it as an opportunity to explore how we could use AI to do better things in a medium that operates with the bare essential staff,” he said. 

“What have we done? Firstly, within OEM, we created a working group with editors from other parts of the country. We started by working on discovering what can be done, what we could use it for, whether there had been any independent local experiences, and what they learned from them.

“One of the first steps with this group was to demonstrate that SEO-optimised headlines generated by ChatGPT are as effective as those created by an editor who knows they will be evaluated. We ran headline tests comparing those created by AI with those generated by expert editors in Chartbeat and found no significant differences in traffic favouring one over the other. 

“Also, it became clear to us that the autogenerative AI is not yet ready to produce reliable texts. The tendency of the model we tested to hallucinate is hard to eradicate. However, we found that by restricting generation only to the data provided, it is an efficient ally and helps us optimise our content to make it more effective for digital media readership. 

“At this point, we want to share the knowledge we have acquired with more newsrooms to try to standardise the prompts and presets so that we obtain similar results across the organisation. Also, we want to train the organisation’s specialised reporters in the analysis of official documents, such as audit reports. 

“Lastly, we have used AI to generate generic images, enhance audio recordings, and improve the quality of photographs. Additionally, we have reached an institutional agreement on how we will use Artificial Intelligence within OEM.”

Small but mighty

By now, we’ve all heard of AI use cases coming out of fantastically innovative news organisations such as Schibsted. But you don’t need to be a giant in the news industry with dozens of different brands to do innovative work. 

Nate McCullough is group editor at Metro Market Media, headquartered in the U.S. state of Georgia, which owns the Gainesville Times, the Forsyth County News, and Dawson County News. He told me: “Our approach could best be described as just getting our feet wet.

“We are a small group of community newspapers. We don’t have the means to do large-scale tech development, so right now we are mainly exploring the use of AI to ease reporter/editor workload by letting the robots do the tedious stuff.  

“So far, we’ve used it for interview transcription, story idea generation, assisting with new products (such as coming up with a name for a growth and development newsletter), and AI-generated voice-over for social media video promos.  

“We hope soon to use it to take content we’ve already produced and repurpose it for special projects.” 

Metro Market also has a committee in place which is drawing up the company’s strategy and rules for AI use. “We even used AI to help generate some of the proposed rules,” he said.  

“One thing we all agree on: It will not replace people here. It will only be used to augment their work and ease their workload.”

Mark Talkington, editor-in-chief at The Palms Springs Post in California, used GenAI to create a tool that could “cover” a Palm Springs Planning Commission meeting and generate a story on one of the agenda items. You can watch his short video on that here. (His newspaper appears to have just two staff members.)

Similarly, Ole Petter Pedersen, editor-in-chief at Norwegian electricity daily Europower, likes AI as a research assistant.

“For my newsroom that covers the electricity market, it will clearly make sense to use AI to read and interpret reports, analyse data sets, keep track of public correspondence, and alert the reporters whenever there is something going on,” he said.  

He also wants to use AI as a fact-checker on the fly during live political debates. And, like McCullough, he is also careful about the output and aware of its limitations. 

“As a research tool, I think we will be able to benefit enormously from AI. But as journalism often is about uncovering what others do not want us to publish, we need to treat it as a research tool more than something that delivers an end product to the user.”

Date for the diary: Inaugural GenerativeAI Initiative Webinar on February 7 

Please join us for our inaugural GenAI Webinar on Wednesday, February 7, 10 Smart Ways to Use GenAI in the News Business.

You’ll hear from two thoughtful speakers on how they are approaching GenAI and the practical ways in which they are using it. We’re featuring use cases from Ringier Axel Springer and DPG Media. Please do join us during this Webinar, which is free to INMA members.

What I’m hearing

• Many news organisations are experimenting with chatbots. This newsletter will have more chat case studies in the weeks ahead. And if you’re looking to build a chat product, please be sure to register for this upcoming Webinar on how to build a chat product for news, run by my wonderful colleague Jodie Hopperton, lead of the INMA Product and Tech Initiative.

• There seems to be a divide in our industry between those who want to move fast and build groundbreaking GenAI products and those who believe there is no real first-mover advantage and want to proceed only after a careful, methodical analysis of use cases. I will be writing more about this in the weeks ahead. Please get in touch if you would like your point of view to be included.

Worthwhile links 

• ICYMI, my colleague Jodie Hopperton chatted with OpenAI’s James Dyett for more than an hour, getting him to answer questions that INMA members were asking. You can watch a recording here and read INMA coverage of it here. 

• What’s next for ChatGPT? “Uncomfortable decisions,” it seems. And if you want even more, here’s Bill Gates interviewing Sam Altman.

• If you’re looking for use cases, here’s a strong piece by Thomas Baekdal on how AI can help journalists

• For journalists working with AI, advice from a reporter who has worked with it for almost a year: “The most powerful results emerge from a series of interactions, not a single Q&A.”

• AI education: Kudos to Jagran New Media for creating videos that teach their viewers about deepfakes, scams, and misinformation (yes, they did use GenAI to produce the videos). I hope my parents are watching them.

• Unlearning: What do we do if we realise that some of our training data needs to be removed after the LLM has already been trained?

• Useful GPTs: The writer complains about useless GPTs (hmm, maybe she should follow our GenAI initiative?); the comments section has good suggestions.

A non-AI diversion

A remarkable natural phenomenon awaits: “There were reports from people in Chicago having to use snow shovels to clear their sidewalks of the dead cicadas.” 

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.

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