We let journalists have fun with AI. They started building solutions.
Over the past year, we designed a new way to teach AI: blending technical understanding with Stanford design thinking and improvisation techniques. After running this approach for 12 weeks at a training program in Europe, we can report: It works.
Here’s what we learned.
This spring, we gathered 17 journalists in Vienna with a hunch: The skills that make great journalists—curiosity, investigation, pattern recognition—might also make great AI builders. We just had to create the right conditions.
The conditions: permission to have fun with AI instead of fearing it. Freedom to try wild ideas without judgment. Space to fail joyfully and call it learning. Room to explore like journalists do best, by diving in and figuring things out.
Twelve weeks later, they were automating website migrations, prototyping news apps, and building tools they actually needed. Most had never coded before.
Having fun at the AI Media Academy in Vienna
The approach: Tech meets creativity
The launchpad for our experiment was an immersive training program that we developed from the ground up in collaboration with the Austrian journalism school FJUM. We called it the AI Media Academy.
We began with the fundamentals. What are LLMs? How do they actually work? We mapped the AI ecosystem and covered the technical realities: probabilistic outputs, hallucinations, biases, copyright concerns.
But understanding alone doesn’t create solutions. So we added Stanford design thinking and improvisation techniques. We played “one word at a time” to demonstrate language models, building up our creative muscles, and started mapping out project ideas. The mantra: Make it exist first; you can make it good later.
Most importantly, we made it fun. When you enjoy the process, you push the boundaries.
What participants built
All 17 participants created remarkable projects. Here are four examples that show the range of what emerged:
Saving hundreds of hours: Evgenia automates tagging
Evgenia Karp from the Vienna-based 1000things magazine faced a massive website migration to a new domain: Nearly 5,000 articles needed recategorizing. Using AI and strategic prompting, she and a colleague automated the entire process. This saved her team hundreds of hours of painstaking work and shows the efficiency gains possible with smart use of this technology.
Date your favorite character: Maya reimagines storytelling
Maya McKechneay experimented with ChatGPT’s new advanced voice mode and saw an opportunity. Her carefully crafted custom GPT called NovelMate lets readers have conversations with fictional characters. You can date Sherlock Holmes and adventure with Dracula. It’s super fun—and in the future, this could be how publishers let readers “meet” protagonists before buying books.
News that knows you: Florian builds an app
Florian Danner, a morning news anchor at the station PULS 4, built a prototype AI-powered news delivery app that dynamically adapts to the user’s mood and preferences. Users can receive updates as text, audio, or through an AI video avatar of Florian himself. When we started, Florian had no coding experience—he built this using AI coding assistants and determination.
Solving a problem: Irene codes a newsletter script
Drowning in AI newsletters, Irene Steindl was looking for a solution. Although she had never coded before, she tried collaborating with Claude to build a Google Apps Script that automatically creates weekly summaries of her emails—and it worked! Now she gets an email digest of her AI newsletters every week, and other participants are asking to get on the mailing list.
Academy participants working on their projects
What we discovered
The biggest discovery wasn’t technical. It was about culture and psychology. Journalism trains us to focus on problems—to investigate what’s broken, hold power accountable, and ask tough questions. This mindset is important, but it can become a barrier when we need to adapt and build for the future.
Journalists already have the core skills for AI work: creativity, persistence, and the ability to learn fast. What we need is the confidence to experiment and build imperfect first versions. Once participants saw that AI could be their partner, everything changed. That’s how they got into coding, even though we never had a dedicated session on it in the program: tools like Claude simply did it with them.
Entrepreneurs and solo journalists became some of the most enthusiastic participants. With limited resources, they saw AI as a force multiplier: They were suddenly able to access capabilities previously available only to large organizations. For scrappy media startups, this democratization holds huge potential.
A final highlight was a one-day startup hackathon where participants used AI to develop and pitch business ideas from scratch. They arrived with nothing but curiosity. 8 hours later, they had frameworks for new businesses, complete with financial planning and marketing assets.
Three short words
The FJUM experiment answered a question we’ve been asking: Can journalists shape AI’s impact on media, or will we just react to it?
In Vienna, 17 journalists chose to shape it. They proved that the future of media won’t be built in Silicon Valley alone. It’ll be built by journalists who understand their craft deeply and aren’t afraid of the future.
So what does that future look like? Nobody knows for sure. But we think it might start with three short words: Make it exist.
Ready to bring this approach to your organization? Get in touch to explore AI training that actually works: [email protected]
Want to join the next cohort? We’re running the FJUM AI Media Academy again this fall in Vienna. Details coming soon.