Natural Language Processing
Administrative processes can be time-consuming for both citizens and public sector employees. Stad Mechelen wanted to make these interactions simpler by exploring how people could communicate with the city in a more natural and accessible way.
Social workers want to help people, not spend hours writing reports. Yet every home visit in Stad Mechelen ended with long, structured documents describing financial, social and personal situations. For many social workers, especially those who only speak Dutch as a second language, this had become a time-consuming and demanding task.
Stad Mechelen was looking for a way to reduce the reporting burden without compromising quality or control.
1
Significantly reduced time spent on report writing.
2
More consistent and higher-quality reports for every case.
3
Fully GDPR en EU AI Act compliant solution with no data storage.
Together with Stad Mechelen, Raccoons built a browser-based voice-to-report tool that lets social workers speak naturally while AI handles the structure.
Social workers simply press record and talk in their own words, in dialect, with local jargon, mixing shorthand, abbreviations and informal phrasing. OpenAI’s Whisper transcribes the speech with high accuracy, further improved by a custom correction agent for domain-specific language.
A second AI agent then organises the transcript into Stad Mechelen’s official reporting templates. The tool never adds or interprets information, it only structures what’s been said. Then reports can be copied directly into existing systems.
The solution fully complies with GDPR and the EU AI Act, keeping social workers in control at all times. Nothing is stored or logged. All processing happens in the browser and disappears when the session ends.
By making reporting as easy as speaking, Stad Mechelen enables social workers to create better reports in less time — and focus on what truly matters: supporting citizens.
.jpg)
%20copy.webp)
