How an innovative AI application, through project-based guidance, grew into an approved quality and risk control measure
Client
Flemish Government entity
Role
Organisational developer - project-based guidance, governance and organisation-wide implementation
Within the GenAI Impact Factory, an experiment emerged in which generative AI was used to support mediators in preparing transmission reports. The experiment showed operational added value: structuring information, making points of attention visible and formulating transmission reports more clearly and consistently. But a promising experiment isn't yet an organisation-wide practice.
Transmission reports are part of a sensitive process context. Quality, completeness and formulation carry weight, but the use of generative AI must not replace the professional judgment of the mediator. The challenge therefore went well beyond the technical quality of a prompt.
A coherent measure had to emerge that supports employees practically, strengthens the quality and consistency of transmission reports, sets clear limits on the use of generative AI, protects personal and medical data, safeguards human control and professional end responsibility, makes risks manageable in advance, and can be tested legally, ethically and organisationally. And all of that in alignment with the broader AI governance of the organisation.
In November 2025, the initiative was converted to a project-based approach. As organisational developer, we guided the transition from an experimental application to a formally assessed, approved and coordinated quality and risk control measure. The bridge was built between the operational expertise of the working group and the broader organisational requirements on quality, privacy, ethics, risk, training and governance.
The application was deliberately positioned as a supporting quality guideline within Gemini. It helps mediators with preparation and formulation, but does not take over professional judgment or decision. The mediator remains the owner of the content and end responsible for the final report. Around that application we built a coordinated framework: clear scope, explicit usage and privacy conditions, quality and validation moments, ethical and legal review, formal inclusion in AI governance and the AI inventory, targeted communication and microlearning, access conditions tied to correct use, and structural follow-up.
The fundamental turning point was the shift from an experimental innovation question to a project-based organisational question. No longer was only the question: 'can generative AI support mediators here?'. The question became: 'how do we turn this experiment into a safe, quality and controllable measure that can be applied organisation-wide?'.
Through that, technology, professional practice, privacy, quality, training, risk and governance were designed as one whole. No longer a standalone solution, but a controlled way of working.
AI experiments don't become sustainable through technology alone. They become sustainable when they are guided in project form and translated into clear roles, quality conditions, risk management, training, decision-making and structural ownership.
That is the core of human-centred digitalisation: not only developing an AI solution, but designing an organisational practice in which employees know what the technology does, what it doesn't do and what they themselves remain responsible for. The organisation now has a controlled way of working with which generative AI can be deployed responsibly within a sensitive client and process context, and a reusable approach for future GenAI initiatives.
