As AI becomes part of the production process in many newsrooms, one fundamental question arises: How can we integrate this technology to strengthen the connection between journalism, formats, and audiences?

During the first half of 2025, I led a workshop at Clarín’s newsroom to explore new narrative formats and embed AI into day-to-day editorial practices. Across four in-person sessions, I worked with more than 50 journalists and editors from multiple sections in a space that combined conceptual frameworks, experimentation, and hands-on content creation.

The goal wasn’t to teach tools but to rethink how information is structured, presented, and even automated — always from a journalistic standpoint. This approach made AI an ally to support and enhance the editorial process.

3 core conceptual frameworks

The workshop was structured around three key frameworks:

1. A user needs-driven approach, based on Dmitry Shishkin’s model developed at the BBC. This framework identifies eight distinct motivations that lead people to consume news: to stay informed, learn something new, solve a practical problem, feel inspired, be entertained, find community, validate their identity, or follow an ongoing story.
2. An analysis of changing news consumption habits. Today, many users access news through search engines or social media rather than a site’s homepage. In this fragmented and fast-paced landscape, news must compete with countless other stimuli for a shrinking attention span. This context calls for storytelling that is clearer, more concise, and modular by design. 

3. A strategic approach to AI as a cross-cutting technology. Rather than replacing journalists, AI can assist, accelerate, and enhance many tasks within the editorial workflow. The goal is not automation for its own sake, but thoughtful integration that supports journalistic quality and efficiency. 

From concept to prototype

The workshop was designed as an intensive hands-on experience, with a structure that differed from traditional training formats. Participants were divided into two groups based on their editorial areas, allowing them to work on formats and challenges specific to their daily production environments.

The programme followed three phases:

• Conceptual: Analysis of news format architecture and modular storytelling.
• Practical: Prompt design and responsible use of generative tools.
• Prototyping workshop: Creation and testing of solutions through design sprints.

The practical phase of the workshop focused on designing effective prompts — understood not as magic formulas, but as strategic instructions that guide AI models to deliver results aligned with Clarín’s editorial standards.

A small but powerful set of tools was introduced:

• ChatGPT: Custom GPTs trained with specific documents and editorial instructions tailored to newsroom tasks. 

• NotebookLM: Used to analyse large volumes of documents and turn them into conversational knowledge bases for editorial reference. 

• Idea starters: A collection of practical examples showing how to integrate these tools without disrupting the newsroom’s editorial workflow. 

AI solutions designed by journalists

Each team developed custom AI assistants to solve everyday newsroom challenges. Four standout prototypes emerged from the workshop:

1. Editorial meeting assistant. Daily planning meetings bring together journalists from different sections to share ongoing stories, but often lack a shared structure or documentation, placing a heavy burden on editors. The assistant developed by the team allows reporters to submit updates via a simple form. It then generates a structured report that organises pitches by section, progress status, and urgency — improving editorial planning and creating a digital log of coverage.
2. Content rewriting for key dates. The Food & Recipes team realised that many high-quality articles go underused when seasonal or commemorative dates arise. They built a custom model trained on past articles to automatically adapt existing content for relevant dates, while preserving editorial tone. This solution boosts productivity and extends the life of archived content with consistency.
3. Contextual linker for ongoing news. In the Breaking News section, the fast pace often leaves little time to search for background. This can lead to fragmented coverage lacking context. The team developed an assistant that scans new drafts, identifies proper names, dates, and key entities, and suggests related articles already published in the CMS. This strengthens narrative continuity and reduces editorial gaps.
4. Automated reader for technical reports. Automotive journalists often need to analyse long monthly reports with detailed vehicle sales data. A custom assistant, trained on past reports and articles, was developed to extract key figures from PDFs and generate a structured draft article — streamlining production without compromising accuracy.

As a long-term resource, a custom model was also developed using all workshop materials and a full list of participants. It allows journalists to ask context-aware questions using natural language, reinforcing learning beyond the sessions and supporting adoption in daily work.

Final reflections

Something deeper is changing. AI is becoming a new interface for accessing knowledge. Conversational chatbots are reshaping how people consume information. Audiences no longer just search: they engage in dialogue. This shift is not merely technological; it is structural.

While much of the news industry sees AI as a tool for boosting efficiency, its real potential does not lie in automation alone. In a landscape of constant acceleration and information overload, its greatest value is enabling a new relationship with time — one that allows newsrooms to reclaim space for analysis and make more deliberate editorial decisions.

Designing a workshop like this required, first and foremost, pausing the logic of immediate content production to create space for reflection.

For many newsrooms, that pause can feel uncomfortable. But one thing became clear: the goal is not to train AI experts, but to empower journalists to regain control over the tools, adapt them to their context, and make them part of their editorial language.

More than adopting technologies, we need to build editorial criteria. The central question is: What tasks could we solve more effectively with the support of a custom editorial assistant? The starting point is not the technology — it’s the criteria: what we do, who we do it for, and how to do it more efficiently without sacrificing quality.

The answers are not in the models; they’re in the journalism.

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