INMA: How Tamedia's internal AI toolbox moved from experiment to daily newsroom use

What the article is about

Nadia Kohler, head of the AI Lab at Tamedia in Zürich, describes how the company built and scaled an internal AI toolbox across Switzerland’s largest newsroom network. The article, published on INMA in April 2026, traces the toolbox from its origins as a small experiment to its current status as a daily resource used by hundreds of journalists and producers across the group.

Context

Tamedia operates more than a dozen publications and employs several thousand people across Switzerland. The AI Toolbox was introduced in 2024 as a controlled environment where a small group of colleagues could test AI tools without risking disruption to existing newsroom workflows. The starting point was a simple internal page with a handful of features. Two years later, the Toolbox is available to all 1,200 editorial employees in English, German, and French.

Key takeaway or method

The toolbox serves three distinct functions in practice.

First, it is a testing environment. New AI tools can be evaluated in a real editorial context before being integrated into the CMS. This allows the organization to assess tools against actual newsroom needs — not vendor claims — before any production deployment.

Second, it provides standalone value for tools that are useful without deeper workflow integration. Kohler cites brainstorming helpers and document converters as examples: features journalists use regularly that do not require connection to publishing systems and that provide direct time savings without process change.

Third, and most relevant for editorial organizations concerned about data privacy, the Toolbox is designed so that use does not train external LLM providers on Tamedia’s content. Journalists can experiment with AI features without the organization’s unpublished reporting, source information, or editorial processes becoming part of an external model’s training data. This was a deliberate architectural decision, not a later addition.

The article’s central lesson is about phasing and trust. Tamedia did not design a comprehensive AI system from scratch and roll it out organization-wide. It built a small, contained space first, let usage patterns and feedback shape what went into it, and scaled from there as adoption proved genuine rather than mandated. The toolbox grew because people found it useful.

Who it is useful for

Editors, digital directors, and newsroom leaders responsible for introducing AI tools to editorial teams. The Tamedia case is particularly relevant for organizations that have faced journalist skepticism about AI adoption, or that have concerns about data privacy when using external AI services.