View illustration full size ↗
The publisher says its in-house AI tools are delivering major productivity gains, as Reach and Reuters develop systems designed to support rather than replace journalists.
Newsroom leaders at Press Gazette’s Future of Media Trends event in London said artificial intelligence is already reshaping editorial workflows, but they stressed that its value lies in augmentation rather than replacement. Christian Broughton, chief executive of The Independent, told the event that editors making full use of the publisher’s in-house AI tools are now “six times more economically productive” than colleagues who do not use them. That claim comes as the publisher says AI now contributes a growing share of revenue and is being folded into everything from summarisation to archive search. The Independent’s AI work began in earnest at the start of 2025 with Bulletin, a product that publishes AI-summarised versions of articles under human editorial supervision.
Broughton said The Independent has built much of its AI stack internally, with guardrails controlling where systems can source information and with final editorial responsibility remaining with humans. He pointed to the company’s 40-year archive as a particular strength, explaining that articles and images were converted into searchable data last year so that the system can surface relevant institutional memory by meaning rather than by keyword. In practice, he said, the tool can alert reporters that the paper has previously covered a subject, provide their earlier framing and offer extra context for a draft, while leaving the decision to include it entirely to the journalist. He also argued that automating technical chores such as tagging and SEO metadata has helped free up capacity for deeper reporting, including recent long-form work such as the podcast investigation SAS: The Shadow War.
Yet the promise of productivity gains sits alongside a broader debate about whether AI truly reduces workloads or simply shifts them. A recent Los Angeles Times report said office workers often spend substantial time babysitting AI outputs, while commentary cited from McKinsey suggested the gains may be felt more by individuals than by companies as a whole. Other industry commentary has made a similar point, arguing that many executives can identify efficiency improvements in theory but struggle to translate them into clear bottom-line results. Against that backdrop, The Independent’s claim of much higher editor productivity will inevitably be read not just as a technology story, but as a question of how much of that gain survives the practical burden of checking, refining and approving machine-assisted work.
At Reach, chief content officer David Higgerson described a similarly cautious but expansive approach. The publisher’s in-house tools include Mantis Scribe, which can rewrite the same story for different local titles, and Launchpad, an editorial dashboard that brings together AI functions, analytics and image access in one place. Higgerson said the starting point is to free journalists for the work that adds the most value, while keeping processes transparent and under human control. He gave one example in which an AI tool working through a 35-page planning document surfaced, almost as an aside, a line about a local café that was due to be demolished. The point, he said, was not that the machine could publish the story, but that it could help the journalist spot what mattered before the item went live.
Reuters executive editor Nick Tattersall said the agency has built tools that suggest alternative headlines and pull out potentially newsworthy facts from source material, but with editors and reporters retaining the final say. He said Reuters uses AI for more foundational tasks only in limited cases, such as automated translation that allows earnings stories to be published in multiple languages almost instantly, and synthetic voice narration on consumer platforms, where readers are told it is AI-generated. He also set out the purpose of Reuters’ new Model Context Protocol server, which is designed to let customer AI systems search and retrieve verified Reuters material using natural language. According to Tattersall, that is intended to prepare for newsrooms in which agentic systems do much more of the background searching, while ensuring they draw on trusted, accurate content.
Marie Bering, vice-president of product management at Stibo DX, said best practice on disclosure is still evolving, particularly for written material that has been assisted by AI. But she drew a clearer line for images and video, arguing that fake visual content ought to be disclosed more plainly. She also said the industry needs better standards, including content credentials that can travel with material through the distribution chain. Elsewhere at the event, Metropolis chief executive Robert Marr said the publisher has responded to falling website traffic by launching ten new apps and seven journalist-led newsletters, while also using AI to tailor recommendations and streamline event production. He said the group’s in-house Miso chatbot has handled 1.5 million reader questions over the past two years, underlining how quickly AI is becoming part of the reader as well as the newsroom experience.
Sources by paragraph: – Paragraph 1: [1], [2] – Paragraph 2: [1], [2] – Paragraph 3: [3], [4], [5], [6] – Paragraph 4: [1] – Paragraph 5: [1]
