Last week, I put out a call for media innovation stories from across the continent, and the response was incredible. I heard from so many newsrooms doing inspiring work, some I hadn't even come across before. Over the next few editions, I'll be sharing their stories.

We recently worked with journalists from The Alkamba Times in The Gambia on the basics of AI. Before the training, I like to do a quick poll to see how people really feel about AI in their newsrooms—is it fear or excitement? Usually, the first reaction is fear, mostly because of worries about being replaced. Sometimes, curiosity sneaks in alongside caution. But by the end of the session, that fear almost always changes to excitement. Watching the Gambian journalists go from anxious to genuinely enthusiastic about what AI can do was an amazing thing to see.

This week, I'm in Uganda, and I've met several journalists who are just as curious about AI and what it means for their work. One of them, who contributes to an international publication, told me his AI-generated story was rejected, and he wanted to understand why. That conversation captured so many of the questions journalists are wrestling with right now.

I also joined the JournalismAI Festival, and it was fascinating to see the creativity emerging from newsrooms around the world. One remark that stood out: do newsrooms really need AI policies if, as one panellist joked, “which no one reads- as journalists are too busy"?

Maybe not fifty-page documents. But I think every newsroom should have at least a short version that clearly states their values and the lines they won't cross.

Take Minority Africa, the newsroom we're featuring today. They've drawn their line clearly: no AI-generated visuals. It's simple, powerful, and says everything about how they want technology to serve journalism, not the other way around.

Across the continent, I can feel the mood shifting. That initial fear has started giving way to curiosity and a real determination to make AI work for journalism, not against it. I think the challenge now is keeping that momentum anchored in strong values, clear boundaries, and a true commitment to doing things the right way.

That’s what The Journovator is all about: to celebrate the people and ideas reshaping African journalism, one experiment at a time.

Can AI Learn Editorial Judgment? Minority Africa Put It to the Test

In this edition of The Journovator, we speak to Caleb Okereke, Executive Editor of Minority Africa. His newsroom is testing whether AI can learn the instincts that make editors human; weighing stories, nuance, and ethics. What does this mean for the future of editorial judgment?

Minority Africa is a digital publication using data-driven multimedia journalism to amplify underrepresented voices across the continent. It covers women, LGBTQ+ communities, ethnic and religious minorities, persons with disabilities, migrants, refugees, and other marginalized groups through solutions-focused storytelling.

The story: Caleb Okereke, Executive Editor at Minority Africa, built an internal AI tool to help editors review story pitches and headlines—trained on the publication’s own archive.

The innovation: Can a large language model (LLM) replicate newsroom judgment? This experiment tests whether AI can sort pitches, offer feedback, and suggest improvements before human editors step in.

Why it matters: With tighter deadlines and heavier workloads, newsrooms are asking: Can AI handle routine tasks without compromising editorial instinct? This project explores that question.

GO DEEP.👇🏾

The Experiment Begins

Every Thursday at Minority Africa, editors gather to review story pitches from their journalists. What should take minutes can stretch into hours.

“Sometimes we spend one hour talking about a single pitch,” recalls Caleb. “We’ve tried so many formats to make the process faster.”

The bottleneck sparked an idea: could AI help without stripping away the human judgment that defines good editing?

“We thought, what if we could work in silos,” Caleb says, “and have something that mirrors our judgment — at least enough to save time?”

That question led to an experiment. Caleb began feeding years of internal data — more than 2,500 archived pitches, each tagged approved, needs development, or rejected, with editors’ written notes — into a custom GPT model. The goal wasn’t to automate journalism. It was to understand it better.

Teaching Machines to Edit

The first version was simple. The team anonymised the data, removed identifiers, and trained the model to classify and comment. Caleb excluded first-person essays to avoid bias.

The AI’s task wasn’t to write stories but to think like an editor — to weigh originality, fit, and focus.

When it worked, it was a genuine time-saver. Editors could run a new pitch through the tool, get instant feedback, and decide whether it warranted a full meeting. Contributors could even test their ideas before submitting.

But the machine also showed its limits. “It overvalued anything that mentioned a minority group,” Caleb explains. “Even when the framing was harmful or clichéd.”

The Friction of Change

When Minority Africa shared the project publicly, reactions were mixed.

“A few people said, ‘I’m going to unfollow you, you shouldn’t be doing this,’” Caleb recalls. “Which is fair — but I actually replied to some of them, saying, don’t unfollow. Tell me what the alternative is.”

For him, that pushback is part of the process. “We’re not having AI review and replace,” he says. “That’s actually contrary to the point.”

He believes the experiment matters precisely because it’s uncomfortable. “We have to build tools,” he says. “It’s different when you build something that mimics editorial labour for a newsroom that’s justice-oriented.”

Why It Matters

Minority Africa’s mission — centring marginalised voices — makes the stakes of automation even higher.

“Just because someone mentions a minority group doesn’t mean it deserves coverage,” Caleb says. “We’re trying to design prompts that make that clear.”

That same principle defines their boundaries. “We don’t use AI to design visuals or create videos,” he says. “It just feels dishonest. Journalism is about truth. If it’s not truthful, is it journalism?”

Still, Caleb insists on engaging with technology critically. “You can only challenge something if you use it,” he says. “There’s no way I’d be able to spot tokenisation or bias if I wasn’t using it.”

Lessons from the Frontline

The experiment has already reshaped how editors work.

“We all got a bunch of pitches to review on our own time, which we never used to do before,” Caleb says. “Having a tool as a partner for ideas makes that easier.”

He plans to refine the model further and present it later this year. “Scaling the data will help,” he says. “But it’s not meant to replace us; it’s meant to be a second opinion.”

For smaller newsrooms, he’s practical: “It didn’t cost us anything to build. You can use a custom GPT — it’s basically no-code. The real work is designing prompts that reflect your newsroom’s values.”

The Big Picture

Minority Africa’s experiment isn’t about replacing editors — it’s about testing the boundaries of what’s possible. By trying to teach machines how editors think, the newsroom is also rediscovering what can’t be automated: empathy, judgment, accountability.

“I still think we need a critical distance from AI,” Caleb says. “It’s not that everything has to be AI or nothing. But we must engage critically and make it work for us.”

For small newsrooms juggling limited resources and big ambitions, Minority Africa’s journey is a reminder that disruption doesn’t always mean replacing humans — sometimes, it means reimagining how they work.

▶ 📢 2027-2028 DW trainee

The application phase for the 2027-2028 DW traineeship has started.

🗓️ Deadline: November 17, 2025.

Apply here 👉 https://akademie.dw.com/en/dw-traineeship-2027/a-70487446

▶ Free online Google course for journalists

Use Google AI tools to improve workflow and engage audiences.

📅 Deadline: running Oct. 20-Nov. 16, 2025.

👉 Apply now: https://knightcenter.utexas.edu/free-online-course-for-journalists-use-google-ai-tools-to-improve-workflow-and-engage-audiences/

▶ 🎙️ Call for Podcasters: Go Multilingual with AI! 🌍

Want to reach new audiences across borders? Join DW Akademie’s free 1:1 mentorship (Jan 6–30, 2026) and learn how to bring your podcast to new languages using AI!

📍 For Sub-Saharan Africa

🗓️ Deadline - Nov 30, 2025

👉 Apply here: https://docs.google.com/forms/d/e/1FAIpQLSfdUHTA-4tD5ROY3QSyUIoNXHXkrSEpBNlfK0cUxTa1Os9Tsw/viewform

We work with media organisations to turn big ideas into real impact — integrating AI, navigating digital shifts, and rethinking editorial strategy for the future. Curious about what that could look like for you? Let’s connect. Send me an email.