The project
The General Secretariat for Catholic Education (SeGEC), which provides services to Catholic schools in Wallonia and Brussels, launched a generative AI project in 2024 to improve the management of teacher leave requests. These requests, which are often complex and frequent, place an excessive burden on the legal department, even though this is not one of its main tasks.
The challenge
The leave circular is over 150 pages long and covers around 120 types of leave, making it difficult for school administrators to interpret. The main challenge was therefore to develop a solution that would facilitate searches within the circular while respecting its structure and regulatory requirements, enabling the legal department to focus solely on complex cases.
Our solution
To address these challenges, Wavenet has implemented a solution based on Azure OpenAI to automate the drafting of responses to teacher leave requests, integrated with other Microsoft services (Azure Document Intelligence, Cosmos DB, AI Search, Cognitive Skills).
Technically, the solution is based on a RAG (Retrieval-Augmented Generation) architecture:
- The LLM (Large Language Model) receives a question,
- It extracts key elements and performs a search in an indexed database,
- Relevant documents are summarized to generate a contextualized response.
Implementation required an iterative Scrum approach, with an initial single-agent phase followed by a multi-agent phase to handle more complex cases (e.g., accumulated leave, which requires specific processing of tables that are not well interpreted by standard LLMs).
Several challenges were overcome thanks to the solution provided by Wavenet:
- Compliance with the regulatory structure of the corpus,
- Management of AI hallucination risks,
- Intensive collaboration with business lines, particularly lawyers,
- Alignment with regulatory requirements (e.g., AI Act).
This project is the first integration of AI into SeGEC. The tool is currently used internally by the legal department, with wider availability planned on the extranet from 2026, in line with the overall strategy of accessibility and automation of services for member schools.
The benefits of this solution are that it provides lawyers with a tool that facilitates searching within the circular. Designed to evolve towards greater automation of responses, the generative AI solution will ultimately reduce the legal department's workload in relation to these requests.
- Implementation of a solution based on Azure Open AI and integrated with other Microsoft services (Azure Document Intelligence, Cosmos DB, AI Search, Cognitive Skills).
- RAG (Retrieval-Augmented Generation) architecture to automate document searches based on a request.
- Second phase involving multiple agents to handle more complex cases.
Let's talk about your project!
Do you have an AI project in the pipeline or already underway? Are you looking to accelerate your transition to data-augmented processes?
Our experts are here to help you identify the most relevant levers for your organization.
