Putting Taiwan's company financials at reporters' fingertips

JournalismAI Skills Lab
Written By Igor Çelov

Felice Fen-Chieh Wu was a senior researcher at a business magazine in Taiwan when she applied to the JournalismAI Skills Lab. Learn how the programme helped her build a financial intelligence tool for journalists covering Taiwanese companies

When a colleague asked Felice Fen-Chieh Wu whether a large printed table of company financials could exist in digital form, she saw an opportunity. As a senior researcher at a Taiwanese business magazine, Wu understood the pain point immediately: journalists covering companies were still relying on static print resources to look up financial data. With a background spanning quantitative research, journalism, and digital transformation consulting, she knew there had to be a better way. The JournalismAI Skills Lab, a programme supported by the Google News Initiative, gave her the structure and support to pursue it.

The solution

Wu's first instinct was to build a natural-language query system using NotebookLM, Google's AI-powered research tool, so that journalists could simply ask questions about Taiwanese company financials and get answers back. But initial testing surfaced a problem she couldn't ignore: NotebookLM's query results were sometimes incorrect which is an unacceptable risk for a tool intended for journalism where accuracy is non-negotiable.

So Wu pivoted. Her final MVP took a different shape: a structured dataset hosted in Google Sheets that lets journalists rank and search Taiwanese companies by financial metrics like revenue and profit margin. The technical workflow combined skills learned during the Skills Lab programme: transforming financial data from the Taiwan Stock Exchange into metrics in a format that reporters could actually use to spot high-performing companies among more than 1,000 companies, or to monitor companies on their beats. 

The tool has clear potential for expansion. Colleagues have already requested features like automated alerts when a company's revenue growth exceeds certain thresholds. “I really do hope that it can evolve into something useful,” Wu reflects.

Key takeaways

Python knowledge remains essential, even when AI writes your code. Wu used AI assistance extensively but emphasises that understanding the fundamentals was crucial. “I can ask AI to do vibe coding, but I still need to check the logic myself. If I don't understand what AI has produced, I cannot properly check its logic.” She considers learning Python essential for anyone building AI tools in newsrooms, acknowledging this might sound radical to some.

Expert support provides a safety net for experimentation. Wu found that having access to mentors who understood both the technical and editorial dimensions of her work made all the difference. The consultation sessions filled a gap she couldn't find elsewhere. “They will be there if you have any questions, so I can just pitch ideas even though I don't know exactly whether it's technically feasible or not.” This confidence boost proved invaluable for pushing through uncertainty, especially for someone who has never built a product before.

Sometimes you just need someone to ask you to build something. For Wu, one of the programme's greatest gifts was simply the structured expectation to create. “The mere request to do something or to build something, and then having the chance to build it – that itself is a wonderful opportunity,” she says. External accountability transformed ideas into reality.

Other JournalismAI Skills Lab Case Studies

Igor Çelov
Next

Scaling local listening: how Radio France used AI to monitor 44 stations simultaneously