News media companies are grappling with many questions related to generative AI, including how to get employees to embrace it and how it can be used outside the newsroom — such as in advertising and marketing.

To help with the journey, INMA’s Generative AI Initiative Lead Sonali Verma brought in two AI consultants for this week’s Webinar: Lukas Görög, a consultant and AI and data strategy lead for NZZ in Zürich, and Justin Kosslyn, special advisor to New York-based Eurasia Group.

A much-needed framework

Görög has worked extensively with companies at various stages of AI adoption and discussed the rapid rise of AI, recalling that transformative moment in 2022 when ChatGPT captured global attention: “That was when we all realised AI is no longer science fiction; it’s our reality,” he said.

That realisation created an urgent need for organisations to create AI strategies. The media landscape, in particular, was profoundly affected. AI was no longer just about text generation; it had become a powerful tool for creating images, music, and videos.

The proliferation of AI tools in the past two years has been overwhelming, making it difficult for companies to determine where to start and which tools to prioritise.

“Now we have so many tools that you basically can spend 24/7 testing them and checking your use cases and it would never end,” he said. “The AI landscape is becoming more and more complicated and it’s very hard to find a proper tool for the task, what we want to accomplish, and what we want to do in our media organisation.”

This widespread confusion inspired Görög to create a structured framework that outlined the stages of AI adoption. His AI Value Framework consists of four key phases: awareness, activation, integration, and operation. Each phase, he explained, represents a different stage of an organisation’s AI journey, and each step increases the value AI can provide.

Step 1: Awareness

The journey begins with awareness, as organisations become familiar with AI and its potential applications.

“At this stage, there are no people actively working with AI,” Görög explained. In this phase, companies can build AI literacy through workshops and training sessions. Companies that invest in AI education for their employees have seen significant improvements in their ability to engage with AI tools effectively.

He listed several key workshop topics that had proven valuable for media companies: AI literacy for journalists, AI in research, AI’s role in combating misinformation, AI applications in podcasts and social media, and data journalism. Through structured training, employees could learn to craft effective AI prompts and understand the ethical considerations surrounding AI use in journalism.

2. Activation

Once organisations have a foundational understanding of AI, they can move to the activation phase, where they begin defining AI use cases and creating pilot experimental projects.

At this point, “AI starts being seen as a tool for experimentation,” Görög noted. Companies can brainstorm ideas on how AI could support their work, testing different applications to determine what yielded the best results.

At this stage, news media companies can develop custom AI bots for content creation, such as generating article summaries, social media posts, and topic suggestions. The key, Görög emphasised, was to foster a culture of experimentation, encouraging employees to explore AI’s capabilities without the pressure of immediate large-scale implementation.

3. Integration

During the third phase, AI becomes an integral part of an organisation’s workflow and it is embedded into systems, Görög explained. AI tools can be incorporated into content management systems, allowing journalists to generate headlines, summaries, and SEO-friendly titles with a single click.

Integration can take many forms, from simple prompt-based features to more complex business logic-driven applications. Some organisations have integrated AI into their internal communication tools, such as Slack or Microsoft Teams, creating AI-driven assistants that streamline workflows.

Others have developed personalised content recommendation systems similar to Instagram’s story ranking algorithms to enhance audience engagement. AI is also used for data analysis, providing media professionals with AI-generated insights on article performance and suggestions for follow-up stories.

Integration must occur within an environment of collaboration with IT and data science teams, Görög said. The most advanced AI implementations involve automating repetitive tasks, optimising content distribution, and personalising user experiences. These integrations not only improve efficiency but also create new opportunities for media organisations to engage their audiences.

4. Operation

The final stage of the framework, operation, is about institutionalising AI within the organisation.

“At this stage, AI needs a seat at the business table,” Görög said.

At this level, companies have dedicated AI teams, such as AI leads or chief AI officers, responsible for overseeing AI initiatives. This phase also emphasises the importance of governance — developing ethical guidelines and transparency policies to ensure responsible AI usage.

Operational AI integration means establishing formalised roles within the organisation, Görög said. Some companies appoint AI ambassadors to advocate for AI initiatives across departments. Others form cross-functional AI teams to oversee strategy and implementation.

As AI becomes a core part of business operations, organisations need clear policies on data privacy, bias mitigation, and ethical AI deployment.

Guiding the AI journey

A common question among news media companies is whether to build AI solutions in-house or buy existing tools.

There’s no one-size-fits-all answer, Görög said: “For the most companies, especially for those smaller media house, it is easier to buy. But if you are big enough and you can afford it, it’s definitely good to try out some building activities.”

Every organisation is at a different stage in its AI journey, he acknowledged, and it’s important for leaders to know where they’re at and where they want to go. “Each stage — awareness, activation, integration, and operation — drives value in a different way.”

Giving AI a job

Kosslyn provided a hands-on demonstration of how organisations can use AI for certain tasks.

“Many folks have spent two years grappling with the basic question, ‘What is this new AI good for?’ And in my experience, what makes this new generation — starting with ChatGPT — different from what came before it is its ability to understand and be responsive to intent.”

Old machine learning systems, he said, “could do amazing things in many cases with enough data” but couldn’t understand the user’s intent or goals: “They were simply mechanical.”

Kosslyn began with a simple example in which he asked a large language model (LLM) to help him build an e-mail newsletter. In this case, he asked the LLM to find “the single best sentence” from an article to use in an e-mail newsletter. He walked attendees through the process and demonstrated how to instruct the LLM, retrieve the requested information, and double-check it for accuracy.

In a second, more complex demonstration, Kosslyn asked the LLM to write a one-paragraph summary of a user’s interest to personalize a news app. He provided the last five articles the user clicked on — up until the place where they stopped reading — and asked what it said about the user’s interest based on which articles they clicked on and finished vs the ones they did not complete.

“You can imagine many uses for this,” Kosslyn said. “Once you start to think about AI through the lens of intent, making sense of a user’s intent, speaking to that intent, you can unlock the full potential and value in my experience of this technology.”

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