By the time this newsletter lands in your inbox, the INMA team will be in New York, putting on the annual World Congress for the 95th straight year (and Earl has probably been there for the previous 94 as well!).

If you’re in town as one of the 600+ attendees and we haven’t chatted yet, drop me a line or come and find me so that we can be sure to connect! 

The positive energy at the event is always contagious, the speakers are first-rate, and the conversations we have are priceless. I look forward to sharing a treasure trove of insights with you once it is over.

In the meantime, please enjoy some best practices from around the world. The first set is from a European media conglomerate that compiled a playbook for media. The second pulls together learnings and insights from various news brands after a year and a half of building conversational interfaces.

Sonali

AI best practices from Ringier

Ringier Media just released a fascinating playbook for digital media, which includes a section on getting your newsroom to adopt AI and also contains many nuggets of insight into best practices and use cases at the Swiss media giant.

Their mid-term goal? Every piece of content should be AI-supported. The reader gets the impression they are ready to roll up their sleeves and do what it takes to make this happen.

Newsrooms should consider AI “a friendly co-pilot” that extends their reach and their effectiveness, Ringier says, and underscores the need to both truly understand journalists’ pain points before building tools as well as to appoint AI ambassadors who foster communication, user buy-in, and ongoing education.

For everyone out there struggling with scaling the adoption of AI tools at their news organisation, the authors of the playbook make it clear that none of this is glamorous work: “It sounds boring, but you have to talk with the real users who will be using these tools in their work,” they say. And also: “A lot of what we are doing at the moment is repeating the message every day.” 

What if it doesn’t work across the board?

Ringier also spelled out in black and white that it is willing to bite the bullet when needed: “There could be a potential need to eventually let go of employees who are resistant to embracing new technologies, especially if their skills are not irreplaceable.”

And at a time when many news brands are struggling with blocking crawlers, Ringier’s view is it is imperative to optimise your content for search generative experiences so that users can discover it in a list of source links while Googling. “Ensuring your content is part of such lists can drive traffic and engagement.”

They also believe readers will prefer to go to news sources over AI overviews and that partnering with Big Tech is vital. “While user behaviours change, user needs do not … . AI chatbots cannot deeply satisfy these needs alone and in many cases are beginning to link sources. To be among the content selected to satisfy these needs, we must build a strong relationship with these LLM platforms, as they act as gatekeepers to visibility.”

Ringier experiments with AI in video and with synthetic voices, and also clones journalists’ voices. “A key point is the focus on high-quality narration, possibly using professional narrators or journalists, rather than solely relying on automated AI voices.”

Having an automated “content distribution engine” that automatically schedules and publishes content with customised promotional copy across different platforms and then tracks its performance is important, the playbook says.

Another place for personalisation: marketing. “Use marketing automation tools to understand user behaviour on your platforms. This allows for target messaging based on user preferences and subscription status.”

Ringier’s North Star is the lifetime value of a customer, and they caution against neglecting subscriber retention rather than simply acquisition. 

When it comes to innovation, the authors believe allocating 80% of resources to the core road map and setting aside 20% for exploring new ideas is wise. 

They also have scathing words for the news business’ current innovation efforts, pointing to silos — rather than integration — as a challenge.

“Media companies are genuinely very good at ‘business as usual’ activities, and they are quite good at innovating on the side, somewhere in the labs, inside some other kind of protected small teams. 

“What they are really rubbish at is actually transitioning the innovation into business as usual.”

Date for the calendar: Wednesday, June 11

I get a lot of questions about print automation from INMA members. Our upcoming Webinar, AI for Automating Print Production, features two speakers discussing their experiences and insights from automating the print production process. Free for INMA members. 

Conversational interfaces: What have we learned?

Chat products are definitely the flavour of the week. Pretty much every media organisation we know has either built one or is thinking of building one.

Why all this excitement about conversational interfaces? What have we learned from building them over the past year and a half? Here are some key insights from different publications in the news business.

• The real utility of a chatbot is in the first-party data you can gather and then use to segment users — and then target them with advertising and offers accordingly, as Jonathan Roberts, chief innovation officer of Dotdash Meredith, points out. Offer the reader prompts that help with segmentation to understand user intent. 

• Use a chatbot as a way of showing value to subscribers. They now have information that is relevant to them at their fingertips instead of having to wade through tens of thousands of articles using poor site search and navigation. As Daniela Buoli, head of product development at Italy’s RCS Media, said when talking about Corriere della Sera’s AI Reader’s Assistant: “People pay for concrete value to solve a real problem.”

• A chatbot can be useful at the top of the funnel as well. You can use it to enrich your content portfolio. Pick a sustainable topic on which you can consistently deliver content on a high frequency, and then produce new content with minimal additional cost, as Ringier Slovakia Communities points out. They found 70% of the readers of their Weather Assistant are new users, and the articles generate 100,000 page views a month. Also interesting: Google did not seem to mind the AI authorship, and Google Discover turned out to be a surprisingly strong source of referral traffic.

• Use it to make complex information more accessible to new audiences, such as younger readers who may not have all the context of an ongoing, long-running story but want to know more without feeling stupid. For example, Norway’s TV 2 ASA created “KI-Kjetil,” a digital avatar of a TV 2 journalist, which users could chat with to learn more about the U.S. presidential election. It was live for 106 hours, in which it answered approximately 70,000 questions from over 42,000 unique users. 

• Use it to reduce the burden on human teams. India’s Amar Ujala created a chatbot to answer audience questions about the Mahakumbh, a Hindu pilgrimage that attracted over 660 million people this year. It handled more than 25,000 queries, reducing dependence on physical help desks, and received an 85% satisfaction rating from users, who cited its quick and helpful responses. Users were guided based on their location and behavioral AI insights. The bot could respond in both English and Hindi.

• Where you place the bot matters. Sweden’s Aftonbladet discovered its audience liked using its Hej chatbot when they placed it at the bottom of article pages. The Texas Tribune saw very little activity on its chatbot when users were asked to go to a dedicated page for it. But after they embedded it on an article page about the chatbot, it gained engagement. One strategy that works: featuring the chatbot prominently in articles optimised for SEO.

• Provide strong prompts/pre-written questions that reflect what your readers genuinely care about. Aftonbladet found this was the quickest way to get some traction, pointing out “the closer it is to the news stream, the better.”

• Scope it narrowly so your site does not get hammered with queries and to make experimentation quicker and easier. For example, The Washington Post created a bot to answer questions on climate change. The San Francisco Chronicle created a chat interface for answering questions on Kamala Harris.

• Consider using evergreen content that will not be overtaken by events, because ensuring a reader gets the most recent news on a topic can sometimes be challenging. For example, Germany’s Funke Mediengruppe created a bot to come up with recipes. Within the first 10 days,  the chatbot recorded almost 4,300 sessions and 600 recipe clicks with a click-through rate of 13%.  

Worthwhile links

• GenAI and theft: Google has been using publisher content for AI Overviews — even if publishers have “opted out” of being used to train Google’s Gemini model.
• GenAI advice: “Treat AI like a smart assistant who tends to be a bullshitter,” says the CEO of The Atlantic, Nicholas Thompson.
• GenAI and scaling: What is more important than AI integration? Audience engagement, according to this survey of 131 media professionals’ priorities.
• GenAI and accuracy: When forced to keep it short, models consistently choose brevity over accuracy and are surprisingly susceptible to the confidence in the user's tone.
• GenAI and accuracy II: An AI-written ad for Coca Cola makes stuff up that is supposed to be factual.
• GenAI and discoverability: How to get your brand into AI answer engines? Talk to journalists!
• GenAI and payments: Perplexity integrates PayPal into its search results for instant direct commerce.
• GenAI and focus: Companies such as Johnson & Johnson are moving beyond the experimentation phase and into the focused application of AI.
• GenAI and the pope: AI developments pose new challenges “for the defence of human dignity, justice and labour,” Pope Leo says.
• An AI diversion: You can take a picture of your Lego bricks and the machine gives you instructions for what to build. Is this a good thing? I think it kind of kills the joy of spontaneous creation.

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.

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