Initiative overview
Public administrations often struggle with a lack of internal technical expertise. This leads to heavy reliance on expensive external IT vendors and slows down the adoption of new technologies. To resolve this, FOD Financiën launched a bottom-up innovation initiative aimed at testing internal AI proofs of concept (POCs) to optimise daily administrative processes. Concurrently, as the subgroup lead for Belgium for internal AI upskilling within the European Fiscalis network, we leverage this local initiative to address a broader opportunity: helping member states achieve AI independence.
The primary objective is to significantly raise AI maturity by empowering government employees to propose, test, and evaluate AI POCs safely. Rather than outsourcing, we focus entirely on building internal expertise and proprietary, reusable assets like our in-house FinGPT platform, data extraction components, and anonymisation tools. A critical, parallel objective tied to our Fiscalis leadership is defining specific internal AI roles and upskilling staff accordingly. This ensures our organisation, and eventually other member states, has the precise competencies required to execute small AI projects autonomously without vendor lock-in.
Over time, what began as a modest POC track managed by four part-time employees has grown into a structured, driving force for digital transformation. We now have a dedicated internal AI team of eleven people operating in an iterative three-month cycle designed to fail fast or scale fast. The internal enthusiasm has snowballed, as evidenced by our most recent call which generated 50 new bottom-up proposals. With 25 tested use cases to date, FOD Financiën currently holds the highest number of tested AI use cases among Belgian federal departments.
In the future, the initiative will be deeply institutionalised through the formal appointment and continuous training of dedicated AI personas across various departments. Because our operational costs remain exceptionally low, with recent POC cycles incurring infrastructure costs of less than 60 euros, we can continuously reinvest in internal knowledge building. The modular nature of our AI components ensures high transferability. We plan to scale our successful POC frameworks and upskilling methodologies to other Belgian government departments and share them across European member states through the Fiscalis network, fostering widespread, vendor-independent AI maturity.
