Processing consultation responses with AI: hundreds of submissions, one response report

By · Updated · 3 min read

Short answer

A draft zoning plan or a large project can bring hundreds of consultation responses, each with its own arguments. AI can sort them, cluster repeated arguments and prepare a draft response per cluster, with a reference to the submission each argument came from. The substantive response and the weighing stay with the authority, because every submitter is entitled to an answer to what they themselves raised.

The pile and the deadline

In the Dutch extended procedure a draft decision is open for inspection for six weeks and anyone may submit a response. The authority must then address every response in a response report. For a zoning plan covering a whole district or a wind farm, that runs to hundreds of submissions, partly identical, partly unique, sometimes dozens of pages long.

Participation adds to it: under the Dutch Environment and Planning Act an applicant must state whether and how the neighbourhood was involved, and the authority itself organises participation for visions and plans. That input has to be processed too.

What AI does, and what it does not

StepWhat AI doesWhat the authority does
SortingSplits each submission into separate argumentsChecks completeness
ClusteringBrings identical arguments together, with references to the submittersJudges whether the clustering holds
SummarisingA neutral summary per cluster, with sourceTests against the original text
RespondingA draft response based on adopted policyThe substantive weighing and the answer
Follow-throughOverview of which arguments led to changesDecides on amending the plan

See Summarising confidential documents with AI for what summaries with source references require.

What to watch for

  • Every argument counts. A unique argument from one submitter must not vanish in a cluster of a hundred standard letters. Have AI flag the outliers.
  • Tone and nuance. A summary must neither soften nor sharpen a submitter’s concern. Spot-checks against the original are needed.
  • Personal data. Submissions contain names, addresses and sometimes medical or financial circumstances. Process them in your own environment and anonymise in the published report.
  • Explainable in court. On appeal it must be clear that every submission was addressed. Keep the link between argument, cluster and response.

How werqly fits in

With werqly the processing runs on a dedicated node: the submissions, with all their personal data, never leave the authority. The audit log keeps which submission belonged to which cluster and draft, which helps on appeal. The same approach works for council questions; see AI for the council and its clerks and AI in permit decisions. See also AI for local government.

Frequently asked questions

Questions on this topic

May AI summarise consultation responses for the report?

Yes, as preparation: splitting, clustering and summarising with source references. The substantive response and the weighing stay with the authority, and someone spot-checks whether summaries do justice to the submitter’s concern.

How do we stop a unique argument from getting lost?

Have AI flag which arguments were raised by only one or a few submitters, and treat those separately. A cluster of a hundred standard letters must not hide an outlying argument.

What does the Dutch Environment and Planning Act say about participation?

An applicant for a permit must state whether and how the neighbourhood was involved and what came of it; the form is free. For visions and plans the authority organises participation itself. Formal consultation responses follow from general administrative law.

May submissions go into a public AI tool?

No. They contain names, addresses and sometimes sensitive circumstances of submitters. Process them in an environment that belongs to the authority alone, and anonymise only when the report is published.

Sources

  1. Informatiepunt Leefomgeving — Participation in permit applications (Dutch)
  2. Dutch General Administrative Law Act, Section 3.4 (public preparation procedure)

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 →