Mid-sized newsrooms like Magyar Jeti’s 444.hu in Hungary are usually very limited in the kinds and quantities of news they can gather. With only 70 employes, the 12-year-old news outlet would not normally be able to assign correspondents to cover the range of stories developing across the world, or trending across social media. AI changes that equation.
Using funding from the Thomson Foundation’s ‘Deepening Digital : Reinforcing Resilience” programme, Magyar Jeti was able to build an AI-powered semi-automated news monitoring system that enabled a dramatic increase in the number of stories that 444.hu’s journalists could access. The early results have been significant, enabling the processing of around 200,000 content items per month and the constant monitoring of 115 news sources across Hungary and the world. This new capability, when fully adopted, will be like adding dozens of new reporters to 444.hu’s newsroom.
A newsroom looking for leverage
Before the project, 444.hu’s journalists did what most mid-sized newsrooms do: monitor key websites, social feeds and wires by hand, one tab and one notification at a time. That work was fragmented, repetitive and impossible to scale.
At the same time, the outlet wanted to broaden its international coverage and react more quickly to stories with relevance for Hungarian audiences – from EU policy debates to culture wars unfolding in the US or political shifts in neighbouring countries.
The answer, for Magyar Jeti, lay in using AI to enable an industrial-scale newsgathering assistant.
Designing an AI-supported monitoring system
With support from Thomson Foundation’s Deepening Digital: Reinforcing Resilience (DDRR) programme, Magyar Jeti embarked on a structured, year-long project with the goal of creating an AI-supported monitoring and story-discovery system that could dramatically expand the newsroom’s field of vision while fitting real editorial workflows.
The team, working closely with external developers and mentored by the author, David Caswell, began by defining which topics mattered most, which domestic and international sources were essential, and how stories should be grouped so they made sense for 444.hu’s editors and readers. This editorial blueprint guided all technical decisions.
The system was developed in an iterative ‘agile’ process in which a ‘minimum viable version of the system’ was built and tested with journalists before gradually adding additional capabilities one-by-one.
These capabilities included: AI tagging and categorisation to understand what each story is about; Domestic and foreign news monitoring to pull in material from Hungary, neighbouring countries and major global outlets; Social media and Reddit monitoring to surface conversations and stories before they hit front pages; Configurable feeds and topic-based views so editors can see exactly the slice of the information space that they cared about; Google Trends integration to spot when a topic suddenly spikes in public interest; Email alerts and newsletters to push the most important developments straight to journalists’ inboxes; and piloted translation services aimed at making international stories accessible in Hungarian.
My role in this project was in providing structured mentorship over the full year of its operation, bringing international experience with AI and product design in major media organisations to the design and development of this system.
As Magyar Jeti COO Gergely László explains: “We held check-ins with David roughly every two to three weeks to discuss the progress of the developments. In the early stages, our conversations focused more on the product itself, its design, and the technical challenges of development. Later on, the discussions increasingly shifted toward the organisational difficulties related to the product’s introduction. In both dimensions, we received highly valuable input.”
The mentoring relationship operated on two levels: as a practical guide contributing to technology and project management, and as a strategic sounding board for bigger decisions. This mentoring guidance often helped the team look beyond individual features and immediate uses and to think about its long-term scalability and sustainability. Gergely added that “Caswell’s insight gave us an added perspective that we would not have been able to bring into the project on our own”
Working in practice
Today, this AI system continuously monitors 115 carefully selected domestic and international news sites – far beyond what any human team at 444.hu could realistically track by hand. It currently processes around 6,700 content items per 24-hour day, or roughly 200,000 items per month.
Each item is automatically tagged and categorised by topic, geography and other editorially defined labels, and enriched with signals from Google Trends and social media. It is then routed into configurable feeds and topic-based dashboards that editors can customise.
What began as a modest plan for a handful of topic filters has quickly grown into at least 14 distinct topic views, including Germany, Poland, the UK, the US, the “last 2 hours”, “international news on Hungary”, and more. These views are already making it easier for editors to filter, prioritise and plan coverage.
As an extension of the project, the team has also piloted a set of curated internal email digests aimed at supporting journalists on specific beats, including a daily newsletter on the United States, in Hungarian, divided into sections for featured news with Hungary/EU relevance, globally significant stories and cultural and social issues. Similar digests now cover trending news from the Middle East and the Far East. Headlines and summaries are translated into Hungarian, while original articles remain in English to preserve nuance and accuracy. This blend gives journalists rapid situational awareness without sacrificing depth.
The tool is currently running as an external system used by a pilot group of four journalists and editors, who explore its functions daily and share rapid feedback with developers in a joint chat channel. All newsroom staff benefit from the summary emails as a new way to track both international and local news.
More stories, less drudgery
Even before a full CMS integration, the impact on editorial work is clear: The newsroom can now keep an eye on “all important media outlets in the world” that matter to its audiences, without assigning people to watch individual sites around the clock; The topic filters, trend data and alerts help journalists identify promising leads quickly, rather than sifting through endless RSS feeds or social media timelines; Automated monitoring, tagging and summarisation free up time previously spent on copy-and-paste tasks and manual tracking; And editors can see how specific themes are evolving internationally and decide when a local angle or explainer is needed.
Instead of replacing journalists, the system acts like a force multiplier. It does the heavy lifting of watching, sorting and summarising, so reporters can concentrate on what only humans can do: verification, context, original reporting and storytelling.
In 2026, Magyar Jeti expects this new capacity to translate into faster and more comprehensive coverage, more relevant stories and ultimately audience growth and stronger subscriptions – all with the same core staff.
Adoption in the newsroom
This technology came together relatively quickly; however the bigger challenge has been making sure that journalists in the newsroom are ready and willing to work in new ways.
From the beginning, the team treated the system not as a finished product, but as an agile and iterative platform to be shaped by journalists and experience. Testing started with the ‘minimal viable system’ early in 2025 with two editors and later expanded to four journalists, whose day-to-day feedback drove iterative improvements. This continuous iteration helped make features more intuitive and ensured that alerts and topic views reflected the real needs of 444.hu’s newsroom.
Even so, fully scaling AI-supported workflows to the whole newsroom will require a deliberate cultural shift. The next phase of work focuses on expanding use in the newsroom by providing editorial workshops and onboarding help for staff, identifying internal “champions” or “ambassadors” to help colleagues incorporate AI methods into their daily routines, and clarification of guidelines on where AI support adds value and where human judgment must be used. This people-centred approach is central to Magyar Jeti’s vision of “responsible AI” – tools that amplify journalistic standards rather than undermine them.
What comes next
With the core system built and early impact visible, Magyar Jeti’s next steps are about deepening adoption and locking in long-term value. The project team would like to integrate the system into their CMS, connecting the AI monitoring tool directly to the editorial system to remove friction. Improvements in translation and enhancements of models and story ranking are also important objectives, as are the provision of personalised alerts to journalists.
In terms of long-term impact, Magyar Jeti intends to relentlessly expand source coverage, adding more European, regional and thematic outlets, as well as deeper social-media monitoring, to catch early signals on emerging issues. The system is now the basis of a strategic roadmap for AI newsgathering, helping to clarify and define how AI will be used across the organisation, including editorial priorities, ethical guidelines, and clear roles for humans and machines.
For 444.hu, the DDRR project has already delivered a step-change in what a 70-person newsroom can see and do in a hyper-competitive information environment. The next phase will determine something even more important: can a pioneering AI newsgathering system becomes an everyday, trusted process in the newsroom – and a quiet but powerful force behind better journalism for Hungarian audiences.
This mentorship process was conducted as part of the Business Innovation Synergizer programme within the Deepening Digital: Reinforcing Resilience project.