NewsLabs builds AI tools for the routine work that fills a newsroom day: turning a wire item or press release into a draft, following up on a social post, transcribing a recording and adapting a story for another format. The company’s promise is speed. Its larger argument is about what a publisher does with that speed once it has it.
Across the industry, AI is often introduced as a way to produce more with fewer resources. NewsLabs challenges publishers to start somewhere else: identify the work that keeps journalists from reporting, give them control over the sources used to produce a draft, and make editorial review part of the system from the beginning. The measure of success, it argues, should include how people work and what the newsroom is able to publish, alongside the minutes taken off a task.
The backdrop makes that argument timely. In the Future Newsrooms Study 2026, FT Strategies and WAN-IFRA surveyed 448 newsroom leaders across 86 countries. Fifty-two per cent cited cultural resistance or scepticism as a barrier to wider AI adoption; 61 per cent cited a lack of technical skills, and 43 per cent expected AI to reduce headcount over the next three years. Journalists have reason to pay attention to how their employers define the technology’s value.
The cost of chasing speed
Plenty of newsrooms are set up to do the same work faster with fewer people. Speed can be worth a great deal when it serves a specific audience need. Reuters’ automated earnings coverage shows where the model fits: automated reporting volume rose 400 per cent, and some stories went out within six seconds. In financial markets, the value of getting structured information to readers quickly is clear.
For a publisher whose value rests on original reporting or specialist judgment, a faster draft is only the start of the calculation. Machine-written copy needs checking. A system that increases the volume of routine articles may also increase the work of verifying them, and newsrooms have to ask whether the saved time produces reporting readers could not otherwise get.
Start with the task
NewsLabs’ starting point is the recurring work that consumes time without requiring much editorial judgment in its first pass: reformatting supplied material, producing a draft from an approved source, transcribing interviews, making summaries for different channels or turning a podcast into a written article. Its platform is built to handle these tasks while leaving the decisions about accuracy, context and publication with journalists and editors.
That approach gives a newsroom a more precise question than whether it should “adopt AI.” It can examine a specific process, the time it takes, the errors it produces and the editorial work that is currently being displaced. A useful deployment removes friction at one point in that process and makes clear who benefits from the time recovered.
NewsLabs says its system reduces the time from brief to draft from 75 minutes to nine in work with its partners. That is a company-reported result. Its significance depends on what happens after the first draft: how much an editor has to correct, how confidently claims can be checked and where the newsroom puts the time it gains.
Design for the editor who signs off
The human approver is a permanent part of the workflow in NewsLabs’ model. A newsroom chooses the wires, feeds or other inputs from which a story can be drafted. Editors can see the source behind highlighted claims and decide what needs further reporting or correction before anything is published.
The BBC’s Style Assist offers another example of a bounded task: adapting reports from the Local Democracy Reporting Service into BBC house style before an editor reviews the result. The value comes from reducing a repetitive rewrite while preserving responsibility for the finished story.
Once editorial approval is designed in from the outset, the review interface becomes central to the product. It must let an editor inspect names, figures and quotations against sources, see what the system changed and make corrections without recreating the original task by hand. The aim is to make judgment easier to exercise consistently across a desk.
That expectation extends beyond publishing. A 2026 Connext Global survey found that 70 per cent of professionals using AI at work associated reliability with AI paired with some form of human review; 64 per cent expected the need for review to increase. For a newsroom, where a fabricated claim can damage both a subject and the publication, the editor’s ability to interrogate the output is a practical product requirement.
What to measure instead
Time saved remains useful, but NewsLabs argues that it belongs beside measures of editorial quality and the experience of the people using the system. The company reports 93 per cent team satisfaction among its partners. That figure is self-reported and needs the same scrutiny as any vendor metric, yet asking journalists and editors whether a tool helps them work is a necessary part of evaluating it.
A study of 543 journalists in China also points to the role of organisational support and communication in effective AI use. A newsroom can buy access to a tool; whether it becomes useful depends on how the work around it is organised.
An AI scorecard should therefore follow the story through the newsroom. How long does the full process take, including review and correction? How often are claims changed or rejected? Do reporters gain time for original work? Can editors trace the final version to its sources? Do journalists choose to keep using the tool once the pilot ends?
Then come the outcomes readers can see. Return visits and completion rates are more revealing than publishing volume alone. So is work the newsroom previously lacked the capacity to produce: reporting from large document sets, useful coverage in additional formats or deeper reporting made possible by time recovered elsewhere.
NewsLabs exists because much of a journalist’s day is taken up by necessary production that leaves too little room for the journalism a publication wants to be known for. Its challenge to publishers is to account for what they do with the hours AI gives back.
Member discussion