AI is reshaping every industry, and newsrooms are no exception. The question is no longer whether newsrooms should experiment with AI but how quickly they can adopt it, and how responsibly they can use it. 

At the recent INMA Asia/Pacific News Media Summit, Lyn-Yi Chung, deputy chief editor at CNA Digital and lead for AI strategy and solutions at Mediacorp News and Current Affairs in Singapore, offered a candid and detailed look at how CNA has embedded AI into its largest newsroom.

For CNA, AI is already powering everyday workflows. One of the clearest examples is AI SmartCut, a tool that automatically cuts TV news bulletins into shorter clips for CNA’s Web site and YouTube. The tool has not only sped up uploads but has also allowed CNA to monetise video content more effectively.

Another headline project is Fast, a feature on CNA’s Web site and app that delivers bullet-point summaries of top stories at the push of a button. 

The summaries are generated with AI assistance but always reviewed by an editor. 

“We rolled it out in March last year, and immediately, the millions more pageviews a month just proved our theory that we have created an experience for people who are time poor,” Chung explained.

The feature has since become habit-forming for audiences, accounting for 10% of CNA’s daily app pageviews. Push alerts for Fast in the morning and evening have driven even stronger results. 

“Sending out the app alert twice a day has led to a 40% increase in daily average PVs for Fast in the app and an almost 100% increase in daily average UVs,” she said. 

For Chung, the lesson is clear: AI isn’t just about efficiency in the newsroom, it is also about creating new kinds of experiences for audiences.

Two years into running an official AI team, CNA has learned what the technology is good at and where it falls short. 

According to Chung, AI excels at monitoring, optimisation, and versioning. “It’s very good at monitoring, and I’m talking about pattern recognition. Flagging events that are possibly newsworthy … I also count transcription as monitoring,” she said. 

The technology also helps CNA tailor social media posts and refine tone of voice, while making it easy to produce multiple versions of a story for different formats or languages.

But Chung was equally clear-eyed about the limits: “It cannot lay on real-world context in real time. That still comes down to a journalist doing the journalism.”

Trust, loyalty, and editorial judgment remain fundamentally human responsibilities: “AI can’t build relationships on its own. It’s still up to us in the newsroom to connect to our specific audiences in different markets and communities … . Loyalty is ultimately the North Star.”

And while AI might sound convincing, it cannot replace the instincts of an editor: “AI itself doesn’t have editorial judgment, even if it’s very convincing and lying that it does. It can’t read between the lines and figure out the nuances.”

Because CNA is a trusted broadcaster across Asia, safeguarding credibility is paramount. The newsroom has drawn firm boundaries around certain uses of AI. 

“We ban generative fill or generative AI to create images and video, and we don’t clone our presenter voices,” Chung said. “Because we are an authoritative broadcaster in Asia, scammers tend to use the likeness of our presenters … to try and get victims on board. So, for now, those things are banned.”

Instead, CNA deploys AI where it can reduce repetitive burdens on staff or minimise human error in tasks like metadata tagging and video logging. 

The aim, Chung explained, is not to replace journalists but to free them to focus on high-impact work. “AI should be the foundation of a story-centric newsroom … . It should unlock value and tailor experiences for our very diverse audiences in different markets.”

Beyond the tools themselves, CNA has also rethought how it organises its AI efforts. Mediacorp runs two AI teams: one organisational unit of data scientists and engineers, and one newsroom-focused team that has recently been paired with Chung’s nearly 20-person growth group.

This hybrid model ensures that the AI strategy is not divorced from newsroom realities. “Half of the growth team is made up of ex-journalists and Web editors, people who care intensely about journalism and want to get the work of my colleagues seen,” she said. 

Sitting slightly outside the daily news cycle, the growth team is well-positioned to test AI tools and measure impact against concrete goals like engagement, loyalty, and traffic.

Chung frames CNA’s projects in terms of “killer” and “filler.” Routine tasks such as writing for social platforms fall into the filler category, while the killer projects — such as high-impact journalism, interactive explainers, and AR storytelling — are what build CNA’s reputation.

By making AI lighten the filler, journalists have more time to focus on the killer.

One of the biggest challenges with AI adoption, Chung noted, is the strain it places on staff during testing. “Testing fatigue is a real concern. You need to make sure that it’s not the same set of small beta testers rating everything … . Make sure you enlist enough people to spread the load,” she said.

To combat this, CNA has woven AI into the structure of its growth team. Members are rewarded for experimenting, sent to workshops and conferences, and even participate in hackathons with partners like OpenAI. This spreads knowledge widely and ensures new tools don’t collapse after their initial rollout.

Chung cautioned against rewarding only flashy launches: “If your system only rewards the rollout and not the implementation and fine-tuning, you’re going to run into scalability problems.”

As Chung sums it up: “We want AI to be transformational in the newsroom, in reducing the burden of journalists … and ultimately, our AI solutions and AI-enhanced systems have to be future-proof.”

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .