Schulink AI-Desk

Municipality of Zutphen

AI assistant supporting municipal professionals in answering legal questions.

Last change on 16th of April 2026, at 8:12 (CET) | Publication Standard 1.0

• Publication category: Other algorithms
• Impact assessment: DPIA
• Status: In use

General information

Theme

Organisation and business operations

Begin date

2026-01

Contact information

[e-mail]

Responsible use

Goal and impact

Purpose: Schulinck AI-Desk was developed to provide municipal professionals (such as consultants, lawyers and policy officers) with instant and clear answers to legal questions. This allows employees to work more efficiently and reach the right solution faster in pending cases.

Impact: In daily use, AI-Desk supports professionals in, among other things, dealing with individual requests for help and securing legal quality internally. This leads to more consistent and careful service provision to citizens, while reducing turnaround time.

Considerations

AI-Desk was chosen because of speed, quality and knowledge sharing. Alternatives such as manual searches or a helpdesk are less efficient. Answers are always based on validated sources.

Human intervention

Schulinck AI-Desk does not make independent decisions. Decision-making always remains in the hands of a human. While Schulinck AI-Desk answers questions, it acts only as a tool for the employee, not as an automated decision-maker. Specifically, a municipal professional uses AI-Desk's advice or answer to support his or her judgement. The employee critically evaluates the output, checks sources where necessary and applies the information to the citizen's specific situation. For example, if the algorithm takes a particular legal position, the official will still assess for himself or herself whether this is complete and appropriate in the context of the case. Deviating from the AI advice is always possible if the situation requires it

Risk management

Only a combination of public and Schulinck-specific content that has been checked for correctness by Schulinck's SME's is used to answer questions. In addition, the answer quality of Schulinck AI-Desk is constantly monitored by Schulinck's SME's and improved if necessary.

In addition, no personal data are processed in the system

Elaboration on impact assessments

The pre-DPIA showed that no personal data are processed. This solution serves as a knowledge base.

Impact assessment

Pre-scan DPIA

Operations

Data

Technical design

Schulinck AI-Desk is based on generative AI with a so-called retrieval-augmented generation (RAG) architecture. This means that the system combines the power of a Large Language Model (LLM) with targeted information retrieval in a proprietary knowledge database. In outline, this looks as follows:

1. Input question
2. A user enters a question into Schulinck AI-Desk in natural language, for example: "What are the conditions for refusing a customised facility Wmo?". The system receives this question text.
3. Searching for sources
4. Schulinck AI-Desk searches in various ways for sources that fit well with the question being asked. The result is a selection of relevant text fragments, for example an article from the Youth Act and an explanation thereof.
5. Answer generation with LLM
6. Schulinck AI-Desk answers the question based on the sources found. In fact, the LLM here functions as a language processor that summarises the legal sources and explains them in understandable terms. The output usually consists of one or more paragraphs answering the question, including legal justification and reference to the sources used
7. Follow-up
8. The user can ask follow-up questions in the same context. So that follow-up questions take the previously found information with them. For example, after a general answer about WMO refusal, the user can ask, "Does that also apply to an application for domestic help?" - the system understands that this refers to the previous question and expands the answer.

External provider

Wolters Kluwer Schulinck