Mediahuis tests AI agents for "first-line" news so writers can focus on signature work
Mediahuis is piloting a system of AI agents to handle first-line news: writing, editing, legal checks, fact checks and multimedia sourcing. Every story still goes through a human editor before publication. The goal is simple-free up 2,000 journalists to do deeper reporting, interviews and distinctive coverage that defines each brand.
Speaking at the FT Strategies News in the Digital Age event in London, head of AI strategy Ana Jakimovska said the early results are promising and the editorial leaders are backing the shift. The bet: routine updates get automated; your best work gets more time.
What counts as "first-line" news
Think short, timely updates with clear public value. Not analysis. Not shoe-leather reporting. Just the fast facts that keep readers informed while the newsroom moves its best people to harder, higher-impact stories.
How the agent workflow is set up
- Commissioning agent: Knows each brand's audience and priorities, scans trusted sources, flags items worth covering.
- Writing agent: Drafts the article to house style.
- Multimedia agent: Finds appropriate visual assets.
- Legal agent: Checks for potential legal issues.
- Fact-checking agent: Verifies claims and highlights concerns.
- Monitoring agent: Watches the reaction. If debate heats up, it sends an opinion/feature skeleton to an editor as a prompt for signature journalism.
- Human editor: Final review and publish.
The data backbone
Mediahuis is building a large, vetted database of sources: parliaments, wire services, think tanks, NGOs, universities, companies and political leaders' official accounts. Examples cited include AFP, Belga, VRT, Oxfam, KU Leuven, the Flemish Parliament, and the European Commission, plus major wires like Reuters.
The commissioning agent pulls from this list, keeping coverage accurate, consistent and aligned with each brand's brief across titles such as De Standaard, De Telegraaf, Irish Independent, Belfast Telegraph, Luxemburger Wort and Aachener Zeitung.
Why this matters for working writers
AI will do more of the repetitive drafting and checks. Your edge becomes reporting, original sourcing, cultural context and judgment. The value shifts to interviews, field work, analysis, and shaping stories that actually move a conversation forward.
If you can brief agents well, triage what's worth pursuing and then deliver a strong human capstone, you'll be in demand. This system doesn't replace taste-it amplifies it.
Practical playbook: collaborate with the stack
- Define "public value". Write a one-paragraph commissioning rule for your beat: what gets covered, what gets ignored, and why.
- Create prompt templates. For fast updates: desired headline format, must-include facts, prohibited claims, tone and attribution rules.
- Source hygiene. Keep a living list of primary sources for your beat and rank them by trust level. Feed only verified inputs.
- Fact-check protocol. Require citations next to each key claim. If the agent can't cite, it doesn't go in.
- Legal red flags. Add triggers for defamation risk, minors, copyrighted assets, and privacy concerns-route to legal agent review by default.
- Multimedia briefs. Pre-approve asset types and licenses per desk; ban ambiguous attributions.
- Monitoring to story upgrade. When polarization spikes, pitch a signature piece: original interviews, data, and a clear thesis.
- Human capstone. Final pass should check: fairness, context, clarity, and whether the piece answers "why it matters now."
What Mediahuis is signaling
The company wants journalists focused on work only humans can do: talking to people, knocking on doors, understanding communities. First-line updates still matter, but they're becoming a system job with strong guardrails and a human at the end of the line.
As Jakimovska put it, this shift could structurally change the newsroom. For writers who adapt, it creates more time to ship distinctive stories-and a clearer path to owning a beat.
Build the skills that complement agents
- AI for Writers: workflows for drafting, editing, QA, and publication.
- Prompt Engineering: structure instructions for commissioning and writing agents that reduce rework.
Bottom line: routine news can be automated. Judgment, sourcing, and voice can't. Lean into the work that only you can do-and use the agents to buy back your time.
Your membership also unlocks:
