PlanGuide

Municipality of Amsterdam

Standardisation and harmonisation of regulations governing the physical environment to improve transparency and simplify implementation in the physical domain (VTH tasks and area development). Furthermore, it is a statutory duty to complete the transition by 2032, and this is not realistic without the use of the PlanGids.

Last change on 2nd of July 2026, at 14:59 (CET) | Publication Standard 1.0

• Publication category: Other algorithms
• Impact assessment: Field not filled in.
• Status: In use

General information

Theme

Field not filled in.

Begin date

2026-04

End date

2031-12

Contact information

[e-mail]

Responsible use

Goal and impact

There is until 31 December 2031 at the latest to consolidate more than 500 local development plans into a single environmental plan. An AI needs assessment was carried out to accelerate the transition and ensure it is implemented to a higher standard and with greater job satisfaction. A number of scenarios have been developed based on this needs assessment, including a comprehensive solution. This comprehensive solution addresses, amongst other things, the possibility of strengthening knowledge-sharing and collaboration, enabling automation in the completion of conversion tables, and ensuring consistency through the reuse of rules. A working prototype, called PlanGids, has been developed and tested, demonstrating that it saves time, that colleagues are enthusiastic about using it, and that it is technically feasible. A minimum viable product (MVP) has subsequently been developed which, following a pilot, can be scaled up to all users.

Considerations

Semantic search may inadvertently reproduce existing inequalities stemming from old rules, and users may be inclined to adopt suggested rules more readily. Furthermore, differences in judgement between planning lawyers and a lack of monitoring may lead to inconsistencies or undesirable patterns in the outcomes.

Proportionality

The use of the algorithm offers clear advantages: it speeds up the implementation of local plans, promotes consistency through the reuse of rules, and supports planning lawyers with targeted suggestions via semantic search. The disadvantages lie primarily in the risk of influencing human decisions and the unintended reproduction of existing inequalities. As the system does not make autonomous decisions and the final substantive assessment always rests with the planning lawyer (and subsequently the planning draughtsman), the use of the algorithm is proportionate and reasonably justified.

Measures

No decisions are made within the PlanGids. The suggestions are still assessed by a planning draughtsman outside the system. This serves as an immediate feedback opportunity to check whether the PlanGids is providing the correct suggestions, or whether the Semantic Search needs to be adjusted.

Human intervention

Through the use of semantic search, suggestions are provided to the planning lawyer for translating zoning plan regulations into environmental plan regulations. The planning lawyer then decides for themselves whether any of the suggestions are appropriate for the existing regulation. If this is not the case, the planning lawyer can draft a proposal for a new rule themselves. This proposal is then assessed by a planning draughtsman, who subsequently enters the rule into another system. No decisions are taken within PlanGids.

PlanGids promotes equality by standardising rules, encouraging their reuse and applying them consistently during the transition to a single environmental plan. Automation and collaboration reduce the influence of subjective interpretations and unconscious biases. Furthermore, transparency and traceability ensure that differences in treatment are transparent and can be rectified, thereby combating unjustified discrimination.

Risk management

1. The PlanGids uses only publicly available data from local planning schemes and the environmental plan.
2. Access to the PlanGids is restricted by personal authorisation based on role or position (only for people working on the translations)
3. The PlanGids only provides suggestions; no decisions are made.
4. A privacy quick scan has been carried out

Operations

Data

Field not filled in.

Technical design

The PlanGids uses public data from local planning regulations. Only the planning regulations (Chapter 3) are loaded into a conversion table. In addition, the most recent version of the Environmental Plan is also loaded. Translated rules are stored in a Knowledge Base so that they can be reused.

Model architecture

1. Zoning plan regulations (Chapter 3 of the zoning plan) are loaded into a conversion table.
2. The most recent version of the Environmental Plan is loaded from a public source.
3. Translated rules are stored in a Knowledge Base

Semantic search allows users to search the Knowledge Base for specific zoning plan rules.

For each search result, the user can see the extent to which it matches the search term (expressed as a percentage). Search results and the translation ultimately selected are logged, enabling monitoring of whether adjustments to settings are necessary.