AI in permit decisions: faster handling, without giving up the reasoning

By · Updated · 3 min read

Short answer

AI can classify permit applications, summarise them and match them against the rules, so the case handler reaches the point sooner. The decision itself stays human work: administrative law requires careful preparation and proper reasoning, and an applicant cannot be dismissed with generated text. Let AI prepare, let the case handler test and decide, and record which source belonged to which finding.

Where the time goes

Permit procedures run on short deadlines. In the Netherlands, for example, the regular procedure under the Environment and Planning Act gives eight weeks, which the authority may extend once by six weeks. Within that period an application has to be completed, assessed and reasoned.

The time is rarely spent on the judgement itself. It goes into searching: which rules apply here, what was said in the pre-application email, which annexes are missing, and what was decided earlier for a comparable building.

What AI can prepare

StepWhat AI doesWhat the case handler does
IntakeClassify the application, flag missing annexesDecide whether the application is complete
Reading the fileSummarise documents, with source referencesCheck the summary against the documents
Finding rulesRetrieve relevant rules from the plan and policyAssess whether those rules apply
Earlier decisionsFind comparable casesAssess whether the comparison holds
DraftA first text for a letter or decisionWrite the reasoning and take the decision

Always have the summary point to the source passage. See AI on your own documents.

Where the limit lies

  • The decision stays with the authority. Administrative law requires careful preparation and proper reasoning. Generated text nobody has checked is not reasoning.
  • No invented rules. A language model answering from memory sometimes cites an article that does not exist. Answers belong to your own current rules. See AI hallucinations.
  • Like cases. If AI surfaces earlier decisions, the case handler must judge whether the comparison holds. Otherwise arbitrariness creeps in, or a pattern nobody chose.
  • Explainability. In an objection or appeal you must be able to show how the decision came about. So log which model used which source.

How to set it up

  1. Start with one permit type with high volume and clear rules.
  2. Measure the lead time and the number of incomplete applications before you begin.
  3. Let AI only prepare at first: summarise, flag, queue up.
  4. Give case handlers access to the sources they need, and no more.
  5. Add the application to your AI register and assess whether a DPIA is needed.

With werqly this runs on a dedicated node in your server room: applications, drawings and personal data stay in your own infrastructure, and the audit log shows who had which document processed. See AI for local government and Generative AI in local government.

Frequently asked questions

Questions on this topic

May a permit be granted by AI?

No, the decision stays with the competent authority. Administrative law requires careful preparation and proper reasoning, and for decisions about people the GDPR limits on automated decision-making apply as well. AI can prepare the work.

Is AI in permit decisions high-risk under the AI Act?

That depends on the application. Annex III covers, among other things, access to essential public services and benefits. An assistant that summarises documents for a case handler usually falls outside it, but a system that assesses or ranks applications may fall within.

How do we stop AI from inventing rules?

Have the assistant answer from your own plan, policy rules and earlier decisions, with a reference to the source passage, rather than from the model’s memory. Spot-check whether the reference is accurate.

What does it deliver?

Mainly shorter lead times in preparation: applications completed sooner, the right rules at hand sooner, a first draft sooner. Measure lead time and incomplete applications beforehand, or you cannot show the effect.

Sources

  1. Informatiepunt Leefomgeving — Regular permit procedure (Dutch example)
  2. AI Act — Regulation (EU) 2024/1689, Annex III

This article is general information, not legal advice. Rules and guidance change; check the sources listed or ask your lawyer or DPO if in doubt.

About the author

Tim Hoogstrate

Founder of werqly

Tim Hoogstrate is a software developer and the founder of werqly. He writes the articles in the knowledge base: what the GDPR, the AI Act and sector rules require when you use AI, and how to arrange that in an environment of your own.

Calculate your private AI costs.

Enter your work email and number of employees. We work out four setups for you straight away: in your own server room or in our datacentre, self-managed or with werqly Pro management. werqly is not a cheap AI subscription: you pay for your own hardware and full control.

  1. 1Enter your email and number of employees
  2. 2Compare four proposals
  3. 3Request a quote straight away

We email the calculation to this address and use it only to contact you about your request.

A no-obligation quote for your own AI node, in your inbox within a minute.

Request a quote →