For small newsrooms operating in crisis zones, maintaining consistent content production becomes a matter of survival. The Ukrainian digital media outlet Babel discovered this reality firsthand when their hypothetical emergency scenario became an urgent operational challenge, leading them to pioneer AI-assisted content creation as a solution for staff shortages.

The problem: The challenge of sustaining journalism during national emergency

In early 2024, Ukrainian media outlet Babel's hypothetical grant scenario became reality: what happens when journalists are mobilised for military service?

"We outlined a hypothetical situation in our submission and then it happened”, the project lead Svitlana Moskalenko recalls. Historian Serhiy Pyvovarov – whose expertise is central to Babel’s YouTube series That's the History – was called up for military duty, leaving the newsroom scrambling to maintain their production schedule with reduced capacity.

So how did Babel leverage AI voice technology to sustain their content production during a national emergency? Here are valuable lessons for small media organisations operating under extraordinary circumstances.

Building the solution: From crisis to innovation

The newsroom's initial vision was ambitious: create AI-powered video content featuring avatars to maintain consistent programming even with reduced staff. "We knew the product we wanted to see. It was in the video format”, explains Moskalenko. Their approach was comprehensive, assembling a toolkit that included:

• Synthesia for avatar creation

• Replit for coding and automation

• Canva for photo and video editing

• Paddle for captions

• Eleven Labs for voice generation

However, the team quickly discovered that not all AI tools are created equal for journalistic purposes.

The Synthesia roadblock: When AI meets editorial reality

The project hit its first obstacle when Synthesia, their chosen avatar platform, refused to process their content due to restrictions on historical and political themes.

This forced a crucial pivot from video avatars to AI-generated voice narration – a decision that proved both challenging and enlightening.

"We need to keep up with production. If we do not publish, we lose our audience," the team realised. The pressure to maintain audience engagement drove rapid adaptation.

Building the team: Skills for an AI-powered newsroom

One surprising discovery was that implementing AI didn't require hiring new technical specialists. "The good thing about AI tools is that you don't need extra skills. You just need to be a little sharp to make it work," notes Moskalenko. The team structure remained largely editorial:

• Editor-in-chief: Provided backup support and helped with text editing

• Designers: Adapted their visual skills to work with AI tools

• Two editors: One for voice, one for video

• Programmer: Handled coding and automation needs

• Remote contributor: The mobilised staff member who could dedicate limited hours

"At its core, it’s still editorial work. We didn’t add anything beyond what’s already part of our daily newsroom agenda,” Moskalenko emphasised, highlighting that AI served to enhance, not replace, traditional journalism practices.

Time, tone, and technology: Unexpected challenges

While the technology worked, implementation brought unforeseen complications:

• Time investment: "We thought AI tools would be easier but they're really time consuming," the team discovered. A single piece required over a full day of editing: "Because of all the names and dates, it has to be generated repeatedly until all the stresses are correct."

• Tonal mismatches: The AI's limitations in matching tone to content created jarring contrasts. "We have feedback that the content is so sad and the voice is so cheerful. So, do something with your host. He's too cheerful for sad things," audience members complained. This challenge was later addressed when ElevenLabs released an update that smoothed out tonal mismatches.

• Financial sustainability: "Most of these programmes aren't cheap," acknowledges the project lead. As the grant ends, the newsroom faces hard questions about viability: "We're thinking about changing the project architecture to make it cheaper and possibly continue without donor support."

The media literacy crisis: An unexpected discovery

The most surprising finding was that audiences couldn't distinguish AI-generated voices from real ones, even with clear labeling. "People don't realise the voice is artificial. We include a notice about AI use, but no one reads it," the team noted.

This highlights broader awareness challenges. As Moskalenko notes: "Media literacy needs to evolve. We assumed everyone could tell the difference [between AI-generated and real content].”

The opportunities: Promise amid technology-literacy concerns

Despite challenges, the newsroom sees potential in their pioneering work. Positive audience feedback validated their experimental approach – "We've received good feedback, and honestly, we expected worse.”

However, concerns remain about the gap between advancing technology and public understanding. "So many tools are evolving quickly, but only a small percentage actually use them," they observed.

Lessons for newsrooms

• Technology evolves faster than implementation: “The pace of technological development is unbelievable. It changes from month-to-month,” the team observed. Tools that were ineffective at the start of the project had become viable options just a few months later.

• AI as agent, not specialist: "AI applications, they're more like agents as opposed to things that require heavy technical knowledge”, making them accessible to traditional newsrooms without extensive retraining.

• Community knowledge gaps: The lack of established knowledge-sharing practices within the journalism community proved challenging: "Most newsrooms tend to work in a fairly isolated way, and there’s no real system for sharing experience within the community. "We didn’t really have anyone in the country we could ask about their experience." This isolation made planning and troubleshooting more difficult.

• Conservative culture as a barrier: “Journalists tend to be conservative, and it’s hard to get them to work with new technology,” Moskalenko noted, reflecting the broader industry resistance to change.