How dmg media is building an AI ‘foundational layer’ for the newsroom
When dmg media set up its Innovation Hub in early 2025, its founding team of four set out to create proof-of-concept products in a sandbox environment to explore potential AI use cases.
Their work produced valuable insights into working with AI, and eventually the team wanted to start scaling up some of these experiments and make them available to the company’s journalists more broadly.
This meant significantly expanding the tools’ user base: UK’s dmg media is the publisher of the Daily Mail, The Mail on Sunday and MailOnline, one of the biggest English-language news websites in the world. Its other titles include Metro, the biggest free daily in the UK, The i Paper, and the New Scientist magazine.
However, implementing proof-of-concepts turned out to be challenging, as “we hadn’t really figured out the route to integrating them within our tech systems,” said Chris Clemo, Director of Innovation at dmg media and creator of its Innovation Hub.
To get support with this step, the company joined the 2025 edition of Newsroom AI Catalyst, a WAN-IFRA accelerator programme in partnership with OpenAI that supports news publishers with their strategic AI initiatives.
Clemo said the team came into the programme with a clear idea of what they wanted to achieve. Internal surveys revealed that journalists wanted to use AI specifically to streamline some of the “periphery, administrative” tasks, allowing them to focus more on writing and generating content.
The goal was to achieve this by building “a powerful foundational layer beneath our workflows, which proactively suggests, optimises and distributes content,” he said.
Elements of Mail iQ
Their solution to this, called Mail iQ, is a suite of tools based on a multi-agentic architecture, built around an orchestrator agent that manages sub-agents handling a range of specific tasks.
These include:
- An editorial style guide assistant that analyses article drafts and provides editing suggestions based on an internal style guide.
- A CMS metadata tool that generates suggestions for SEO headlines, tags, URLs and other fields that need to be filled out when journalists file an article.
- An analysis layer which provides actionable insights based on historical internal performance data, and identifies external live trends and suggestions based on real-time social media content.
- A social asset generation tool that creates social media content based on an article, for the social media team to review before posting on social media platforms.
- A newsroom intelligence feature that incorporates the above points and provides journalists with relevant suggestions as to what timely content to add in their articles, as well as additional agents that are being continually added to the Mail iQ system.
The overall ambition is for Mail iQ to become “the foundational layer that powers all of the different teams,” he said.
Crucially, Mail iQ’s agents do not generate new journalistic content as such. Instead, their outputs are based on journalists’ own writing. Moreover, its suggestions always need to be validated by humans.
For example, the social asset generation tool looks at information included in an article (images, headline, copy) and generates assets based on dmg media’s own social media guidelines. While these assets are publication-ready in principle, they still need to be reviewed by social media editors before posting.
A diagram of the Mail iQ agentic system.
Boosting social distribution
The long-term goal is to implement all the different features into dmg media’s CMS, which the company has developed in-house, giving all journalists an easy access to them.
The newsroom intelligence feature, for instance, exists as a Chrome extension that plugs into the current CMS, while the social media tools and editorial style guide assistant are separate apps.
Thanks to this implementation, the social tool is now available to the company’s social teams in the UK, the US and Australia. They are actively using it to create more than 300 assets every day.
Mail iQ’s Social asset generation tool
“The social teams absolutely love it,” Clemo said. “Before, it was taking them, say, 5 minutes to post something, and now it takes them less than a minute.”
This not only allows them to post at a higher volume but also makes it easier to post on different platforms more strategically. For example, they weren’t posting much on Reddit before, but now this is much easier as the tool suggests appropriate subreddits to post articles to.
AI editorial assistant
As for Mail iQ’s other features, most of these are currently used in the US, Clemo said, as the company’s US office often champions new tools before they are rolled out more widely.
For example, the metadata field suggestor is being used by the whole US office as well as some of their UK colleagues, while the editorial style guide assistant tool is being used by a third of the newsroom globally.
The latter tool has been particularly effective in helping maintain a uniform and consistent quality across the copy from the different journalists using it.
“It is very hard to get the newsroom in three continents working under one style guide,” Clemo said. But if you train an LLM with your guidelines – from whether to use percentage symbols, to when numbers should be spelled out – it helps you maintain uniformity across the brand, he said.
“That’s been really helpful, and the feedback has been great. All they want is it to be integrated in the CMS. And we’re working actively to get that to happen,” he said.
Editorial style guide assistant
Getting everyone on board
The priority for the Innovation Hub team – which recently hired two additional AI engineers, bringing the total number to seven – is now to fine-tune the existing tools and keep developing them in parallel to the on-going development of the new CMS.
The team is developing the tools to be “CMS-agnostic,” so they can eventually be integrated into any CMS and used by other titles within the group (Metro, The I Paper and New Scientist).
Further development and adding new features is fairly straightforward, thanks to the multi-agentic framework that powers Mail iQ. Now that the foundations are in place, adding new agents takes hours rather than days, Clemo said.
He added that eventually there will be new tools available to journalists: “We’ll build different agents and tools for the different teams based on their requirements.”
The team from dmg media that took part in the Newsroom AI Catalyst programme.
