Together with the NVWA Innovation Lab we built AGORA, a proof of concept for an AI-driven multi-agent system in supervision. Where most AI tools ease a single task, AGORA brings several specialised agents together under a central orchestrator.
The user puts a question to the orchestrator, which automatically calls in the right agent: retrieving regulations, analysing documents or drafting a report. Every answer points to the source it came from, so the user can check what it is based on.
In this first application, AGORA supports inspectors in preparing, carrying out and wrapping up inspections. People stay in charge and stay responsible: human in the loop is a starting point, not an afterthought.
We built two equivalent variants, one on open source and one on closed source, and compared them on reliability, future-readiness, security and cost, tested against frameworks such as the GDPR and the EU AI Act. That turned the technology choice into a well-founded trade-off rather than an assumption.
The outcome: the multi-agent architecture is feasible and scalable, with an estimated 40 to 60% reduction in administrative burden at wider deployment. As a proof of concept, AGORA thereby lays a foundation for responsible AI support within the NVWA.
