Are you one of the many people who are dismayed by the U.S. election result or are increasingly grumpy over the inexorably shortening days in the northern hemisphere? 

Never fear — today’s newsletter is here to give you something positive to chew on, full of constructive insights about our industry, particularly in the field of GenAI.

Today, we bring you the highlights of a wide-ranging report on how media companies in the UK are approaching GenAI. And we are also turning the spotlight on one particular U.S.-based news publisher that is using GenAI for its advertising and marketing departments (because many INMA members told us they are particularly interested in those use cases). 

So, please, feel free to share this e-mail with your colleagues and ask them to sign up to receive more like it in the future! And if you would like to offer feedback on what you would like to see in this newsletter, you know where to find me.

Sonali

GenAI experiments in the UK: What’s new

Britain’s news brands are experimenting widely with GenAI, mainly to find efficiencies but also to serve their audiences better, according to a report by research and advisory firm Enders Analysis. 

The report talks extensively about the risks of AI-generated content, which I believe the readers of this newsletter are already acutely aware of. So instead, I’ll focus on the key findings that may be news to you: 

• Using GenAI to cut costs will not offer any business a competitive advantage because all businesses can access these tools. “But there are opportunities to improve service, reduce costs, and potentially develop new products that could build out utility to readers in the context of a shift to digital reader revenues — even if there is no immediate, killer news use case to raise new revenues,” wrote authors Niamh Burns and Tiffany Chung.

• New product development is in the very early stages for UK publishers. “Some are dipping a toe in the water with ‘chatbot' interfaces on their sites,” which could pave the way for improvements to their sites and let them extract more value out of their archives, which are generally not well monetised.

“Products could be developed that allow users to discover more of this,” the report said, citing Condé Nast’s partnership with Google to digitise Vogue archival material as an example. “AI can assist in generating more sophisticated metadata for archival material that would have been far more labour-intensive before and that has been a barrier to better use of the archive (this is a stated goal of the BBC using generative AI, building on its previous COMMA project to make the archive searchable).”

• Generative AI doesn’t represent a revolution in the newsroom — but combining it with rules-based AI and human intelligence can be powerful. For example: The Financial Times may use LLMs to classify information in large datasets but then use traditional rules-based AI to perform analysis on that data. 

• Newsquest has built a range of custom tools it is deploying across its newsrooms — like a freedom of information (FOI) bot that can generate story leads through automated FOI requests and identification of newsworthy responses. 

• Newsquest also encourages its newsrooms to create their own applications based on a GenAI toolkit.

• The British Broadcasting Corporation creates synthetic audio forecasts for every postcode in the UK. The synthetic voice was created by recording 3,000 sentences to create a natural-sounding voice clone, which can read out a weather forecast, a news article or any other BBC data, or text. The BBC is also trailing adding generative AI subtitles to podcasts on BBC Sounds.

• FT Professional has tried an “AskFT” chatbot format with some subscribers, finding relevant articles in response to user queries. FT Professional subscribers can also highlight text in an online article and are offered other articles semantically related to the highlighted section. 

• The Telegraph, The Guardian, Reuters, and News UK all use GenAI for “back-office” tasks, such as headline suggestions, editing help for journalists, transcription, story ideation, and research assistance.

• Reach Plc’s “Gutenbot” modifies existing Reach articles for different publications, as well as transforms press releases or police reports into copy. Reach also uses GenAI to write articles and says it produced 6,000 articles “with the support of AI tools” in 2023.

• Future Plc uses GenAI to generate product specifications for articles, shaving 30% off writing time, on average. It also acts as a co-pilot for editors and edits video for different distribution channels. Future was one of the first UK sites in March 2023 to launch a chat product for users to interrogate for its Tom’s Hardware brand.

The report also offered business strategy advice:

• “New product offerings could be more of an opportunity for businesses that rely on subscribers than those that are ad-supported.”

• “The best practice we identify involves aligning experiments with business strategy.”

• And as for bad practices, such as using GenAI to generate content at scale: “You should ask if you have a business problem that will be solved by more content, especially if that content will stand out less and less online — or by cheaper content, at the expense of a reduction in quality.”

GenAI Town Hall: Did you miss our look back at what we have learned in 2024 and our glimpse into the future? You can watch the recording here.

Reminder: Are you working on something interesting? I am always on the lookout for GenAI use cases to feature in this newsletter and on Webinars. Please get in touch! 

GenAI in advertising: Hearst creates solutions

Hearst Newspapers is using GenAI as an assistant for its advertising teams, helping them build stronger connections with prospects and customising their initiatives from the first cold call all the way through the proposal presentation and campaign execution.  

The U.S.-based news chain, which owns brands such as the San Francisco Chronicle, found ad sales were becoming increasingly complex. 

“E-mailing prospects and developing media plans used to be more straightforward if you look back 10 to 15 years ago,” said Joel Laffer, senior vice president of marketing operations and advertising strategy. “It was: Are you looking to get more new clients? We represent the newspaper in your town, and we’d love to work with you.” 

But now, he said, news brands need to prove their value in a far more competitive and complex marketing landscape. “You have to bring industry expertise and new ideas that are specific to your clients’ challenges to even get their time. And how do our sellers have time to become experts in everyone else’s business? It's really difficult.”  

Michael McCarthy, senior director of AI, sales, and business solutions, used GenAI to build tools that prepare his team for sales conversations, helping them ask “smart, open-ended questions to accelerate the sale considerably.” These include personalised coaches sales reps can talk to, preparing them for client interactions, as well as creating agency-level proposal presentations. 

“Calibrating a suitable multi-channel campaign after a conversation is also hard because the portfolio of products we offer is extensive, and each customer has specific goals and preferences that our sales reps need to account for as they tailor their specific media plans,” McCarthy said.

“That’s actually a very complicated formula to figure out because there’s no one right answer. It’s the right answer for the right situation at the right time,” he said. “Our team needs expertise in dozens of variables, including the specific industry, target audience, digital and print product offerings, competitive landscape, campaign objectives, etc.  What if we could scale that expertise and democratise it? AI, we found, can help us do that.” 

Hearst also uses GenAI to draft “personalised and engaging e-mails” that contain language tailored to show prospects its unique value proposition specific to their business. It is also using GenAI for creating better content for its advertisements for a high volume of clients at scale, Laffer said: “Our ambition is to offer clients, at all budget levels, creative optimisation that resonates with our readers and drives the best possible performance.”

Hearst’s GenAI tools mean new sales hires no longer necessarily need to come with media backgrounds. 

“You’re looking at a different skill set now,” McCarthy said. “It’s more about, ‘Hey, do you have the drive and the soft skills to communicate well with people? Are you intellectually curious about marketing? Can you translate data and insights into a compelling marketing strategy and be a valued partner to our clients?’ Nine times out of 10, AI can help teach them the product and industry background needed to set them up for success.”

Worthwhile links

• GenAI and copyright: OpenAI broke the law, according to a former employee.
• GenAI and copyright II: Ziff Davis says high-quality publishers’ content is disproportionately used to train LLMs.
• GenAI adoption: Looks like many companies are struggling to scale GenAI adoption after running experiments, something we have heard from many INMA members as well.
• GenAI adoption II: Half of all desk workers would be uncomfortable admitting to their manager they used AI for common workplace tasks for fear of being seen as less competent or lazy.
• GenAI and hallucinations: LLMs often present lies even when encoded with truthfulness tokens.
• GenAI and hallucinations II: Here is how the UK government is warning users of its new chatbot about possible GenAI lies.
• GenAI and hallucinations III: Teaching an LLM to think more like a human reduced inaccuracies.
• GenAI and covering up: Most writers who use GenAI do not tell their manager, survey finds.
• GenAI and summaries: Lost the plot? If you’re streaming, Amazon Prime will recap the story for you.
• GenAI and bias: Much of the world is not represented in the data used for LLMs.
• GenAI and advertising: Retailer Mango is using bots instead of real models, as are Nike, Louis Vuitton, and Levi Strauss & Co.
• GenAI and energy: Your electricity bill may go up.
• A non-AI diversion: How long is a day? It looks like man-made structures are affecting the earth’s rotation. 

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.

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