WAN-IFRA: Ezra Eeman on where newsrooms stand in the shift to agentic AI
What the article is about
This article by Ezra Eeman, WAN-IFRA’s lead on AI in media, surveys the current state of AI adoption across newsrooms. Eeman distinguishes between AI use that is already standard practice and emerging workflows that represent a more significant change in how content is produced and distributed.
What it covers
The article identifies the tasks where AI is now routine in many newsrooms: transcription, generation of social media variants from existing articles, and verification assistance tools that check quotes, dates, and facts against source material. These uses share a common pattern — AI acts on content that journalists have already produced, rather than initiating new content or making editorial decisions.
The more significant territory Eeman describes involves multi-step agentic pipelines. Two cases illustrate the direction. TNL Media Group’s Genie system is described as an agentic newsroom: a set of connected AI components that can move an article through stages of production — from a source brief to a draft, through editing and formatting — without manual handoffs at every step. Mediahuis has built a pipeline where a draft generated from source material passes through sequential agent steps: editing, fact-checking, and legal review, each handled by a separate agent component before a journalist sees the output.
Eeman introduces the concept of “news atoms” — individual facts, quotes, and data points extracted from reporting, tagged for machine retrieval, and stored in a form that AI systems can reassemble into different article formats for different platforms or audiences. The concept represents a shift from writing articles to producing structured data that generates articles.
Key takeaway
Eeman is explicit about the current limits: “real autonomy is still very much an illusion.” The agentic systems in production require substantial human oversight and are not fully autonomous. The article functions as a map of where practical adoption is, rather than a forecast of where it is heading.
Who it is useful for
Journalists, editors, and media technology leads at organizations that are evaluating or building AI production pipelines. The article is most useful as a reference point for comparing an organization’s current state against what others in the industry have deployed, and for identifying which specific pipeline components are mature enough to adopt now.