AI in municipal debt counselling: early warning, intake and plans, without risk scores

By · Updated · 3 min read

Short answer

Since 1 January 2021 Dutch municipalities must act on payment-arrears signals from landlords, energy and water companies and health insurers and offer help. That is a lot of work with highly sensitive data. AI can organise the file, write up conversations, prepare a plan of action and make letters understandable. What AI should not do: compute a risk score per resident. That is exactly where SyRI and the Rotterdam model came to grief.

The task since 2021

The amendment of the Dutch Municipal Debt Counselling Act made early warning a statutory task. Fixed-cost partners report an arrear, the municipality must assess the signal and make contact with an offer within four weeks. In a mid-sized municipality that means thousands of signals a year, some stacked from several partners, followed by regular debt counselling: intake, stabilisation, a debt arrangement or administration.

Everything in it is financially and personally sensitive: income, debts, health, family. The people involved are in a vulnerable position, which both the data protection authority and the AI Act weigh heavily.

Where AI helps

StepWhat AI doesWhat the counsellor does
SignalsBundles signals from several partners at one address and shows what is already runningAssesses and chooses the approach
ContactA letter or message in plain language, tailored to the signalCalls or visits
IntakeWrites up the conversation and fills in the overview of debts and income from documentsChecks and completes
PlanDraft plan of action from the intake and the policy rulesSets the plan with the resident
CreditorsPrepares proposals and calculations, organises correspondenceNegotiates and decides

See Plain-language letters with AI: a letter someone understands raises the response to a signal more than any model.

What AI should not do

  • Risk scores per resident. A model predicting who will get “problem debts” works on exactly the characteristics that led to discrimination in SyRI and the Rotterdam welfare model. Prioritise on the nature of the signal, not on a profile. See Testing algorithms for discrimination.
  • Deciding. Admission to debt counselling is an administrative decision. A person takes it, with reasons.
  • Sharing data that may not be shared. The Act sets out precisely which data the municipality may request. An AI assistant that “helpfully” gathers everything goes beyond that.
  • Public tools. An intake report with income, debts and family situation does not belong in a tool outside the municipality.

See also AI in social services for the broader line.

How werqly fits in

With werqly the processing runs on the municipality’s own node. Debt counselling gets its own role in werqly AI Control with only its own sources: policy rules, templates and the counsellor’s file. The audit log shows which document was used for which plan, which is what the council and the supervisor want to see. See AI for local government.

Frequently asked questions

Questions on this topic

May AI decide which early-warning signals we follow up?

Bundling and organising is fine: several signals at one address, what is already running, which deadline applies. A risk score per resident based on personal characteristics is profiling that is legally vulnerable and has led to discrimination in practice.

Can AI write up an intake conversation?

Yes, with a transcription in your own environment and a summary the counsellor checks. The conversation covers income, debts and often health or family; that must never pass through a public service.

What deadline applies after an early-warning signal?

The municipality must offer a first conversation within four weeks of receiving the signal. AI helps meet that deadline by bundling signals and preparing the first contact.

Is debt counselling high-risk under the AI Act?

Annex III lists AI deciding on access to essential public services and benefits. A system that determines admission or priority may fall under it; an assistant that prepares reports and letters does not.

Sources

  1. Dutch Municipal Debt Counselling Act (Wet gemeentelijke schuldhulpverlening)
  2. Government of the Netherlands — Municipalities: earlier signalling of debts (Dutch)

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 →