Smart processing of WOZ Objections

Municipality of Rotterdam

A digital assistant to support the appraiser. The assistant works by reading and processing letters of objection and evidence and helps prepare response letters.

Last change on 2nd of December 2025, at 15:28 (CET) | Publication Standard 1.0

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

General information

Theme

Living

Begin date

03-2025

End date

12-2025

Contact information

[e-mail]

Link to source registration

https://algoritmeregister.rotterdam.nl/p/Onzealgoritmes/53761720552094553

Responsible use

Goal and impact

PURPOSE: The goals underlying the desire to partially automate this process are:

1.Meeting statutory deadlines for processing WOZ objections by speeding up.

2. Shortening processing times and reducing waiting times for citizens.

3. Reducing amount of screens needed to process objections

4. Working more efficiently and reducing error margins

IMPACT: The information from the supplied documents is read by the application and prepared in the programme. Here, the documents are read and linked to each other so that the handling process can be accelerated. After an employee carries out the assessment process, the assistant helps prepare a letter with the known information. Employees check the letter and after this check, the results are saved and the letter is sent.

Considerations

The advantage of using this application is that the process can be significantly accelerated. So that it becomes clear quicker what the ruling is going to be or whether more information is needed, for example. Employees of the Municipality of Rotterdam spend less time retyping information or attaching supporting documents. The objection assessor always remains in charge and has full overview of the process, but now focuses on assessing instead of administration.

Disadvantages are that personal data is temporarily stored and processed in a cloud environment. This is minimised by not using sensitive data and removing data from the cloud environment as soon as possible. There is a small risk of errors by the algorithm due to the large variation within applications. For this, an explicit check by an employee is built in on each step.

Human intervention

The algorithm application remains an assistant and only supports the appraisers in processing the objections. The appraiser is in control at all times and assesses all information including findings from the assistant. Also, the last step in the process, namely adjusting the WOZ value (if any) remains completely manual. Furthermore, based on test samples, cases processed with the help of the assistant are subsequently checked by staff members.

Risk management

For each case, the outcome is checked by a colleague. This process of checking is reviewed quarterly.

Legal basis

Property Valuation Act

Links to legal bases

Property Valuation Act: https://wetten.overheid.nl/BWBR0007119/2025-01-01

Elaboration on impact assessments

DPIA: Privacy risks are considered low for this process and applying the smart virtual assistant

Impact assessment

Data Protection Impact Assessment (DPIA)

Operations

Data

Technical design

First, the grievance document is automatically converted to text readable by the programme. Then, parts of this text are linked to known categories of grievances through an AI model (Azure Cognitive Services - Text Embeddings). Any included evidence such as photographs are linked to these found grievances through image recognition. While taking these steps, possible 'false positives' and 'false negatives' are also closely monitored.

The pilot phase examines how often these errors occur and under what circumstances. After an employee has checked these above steps of the application, the objection is processed. In doing so, all information is now in a logical place, allowing for faster assessment.

After processing all the grievances found and entering any new WOZ value, the algorithm automatically completes the response letter. This response letter is not a product of generative AI, but follows a fixed format as determined by the template. The template is filled in with standard texts. Thus, in principle, the employee no longer has to cut and paste, but only has to proofread the letter in most cases.

If anything is wrong, the employee can easily correct it. After a final approval by the employee, the letter is automatically sent to the citizen and the rest of the information is transferred to the municipal systems.

This is better than the current situation without the use of this application. Because a lot of time is saved by the municipality's employees. Residents of the Municipality of Rotterdam will not have to wait as long for a ruling or a request for additional evidence, so they will have clarity faster.

Furthermore, the municipality will be able to gain better insight into the process surrounding WOZ objections, making future improvements easier to realise. In addition, insightful information (e.g. common grievances) is not currently stored. Thanks to the link to known grievance categories, this information is collected in a structured and anonymous way for future analysis. Regular samples are taken and there is human supervision of every step of the process. So that the quality and reliability of the programme is guaranteed all the time.

External provider

NextHuman B.V.